Amorphous Semiconductors for Gamma Radiation Detection (ASGRAD)

PNNL-22277 PL06-136-PD05

FY08 Annual Report: Amorphous Semiconductors for Gamma Radiation Detection (ASGRAD)

B. R. Johnson B. J. Riley J. V. Crum J. V. Ryan S. K. Sundaram J. McCloy A. Rockett, UIUC

February 2009

DISCLAIMER

This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor Battelle Memorial Institute, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof, or Battelle Memorial Institute. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

PACIFIC NORTHWEST NATIONAL LABORATORY operated by BATTELLE for the UNITED STATES DEPARTMENT OF ENERGY under Contract DE-ACO5-76RL01830

This work was conducted in part at the Environmental Molecular Sciences Laboratory (EMSL), which is operated by the Office of Biological and Environmental Research of the U.S. Department of Energy (DOE) at Pacific Northwest National Laboratory (PNNL).

Printed in the United States of America

Available to DOE and DOE contractors from the Office of Scientific and Technical Information, P.O. Box 62, Oak Ridge, TN 37831-0062; ph: (865) 576-8401 fax: (865) 576 5728 email: [email protected]

Available to the public from the National Technical Information Service, U.S. Department of Commerce, 5285 Port Royal Rd., Springfield, VA 22161 ph: (800) 553-6847 fax: (703) 605-6900 email: [email protected] online ordering: http://www.ntis.gov/ordering.htm

PNNL-22277 PL06-136-PD05

FY08 Annual Report: Amorphous Semiconductors for Gamma Radiation Detection (ASGRAD)

B. R. Johnson B. J. Riley J. V. Crum J. V. Ryan S. K. Sundaram J. McCloy A. Rockett, UIUC

February, 2009

Work performed for the Office of Defense Nuclear Nonproliferation (NA-20) Office of Nonproliferation Research and Development (NA-22) under Project Number PL06-136-PD05

Prepared for the U.S. Department of Energy under Contract DE-AC05-76RL01830

Pacific Northwest National Laboratory Richland, Washington 99354

Summary

High Z, high resistivity, amorphous semiconductors were designed for use as solid-state detectors at near ambient temperatures; their principles of operation are analogous to single-crystal semiconducting detectors. Compared to single crystals, amorphous semiconductors have the advantages of rapid, cost- effective, bulk-fabrication; near-net-shape fabrication of complicated geometries; compositional flexibility; and greater electronic property control. The main disadvantage is reduced-charge carrier mobility. This project developed amorphous semiconductor materials for gamma detection applications that leverage material advantages while mitigating the limitations of this class of materials. This development effort focused on materials synthesis, characterization of physical properties, characterization of electrical properties, radiation detector design, as well as detector signal collection and processing.

During the third year of this project, several important milestones were accomplished. These include: 1) We were able to collect pulse-height spectra with an amorphous semiconductor-based radiation detector. This was done using both an alpha source and a gamma source. 2) We identified an anomalous conduction phenomenon in certain amorphous semiconductors. When cooled to approximately 250K, they displayed enhanced conductivity by several orders of magnitude. This increase is thought to be due to increased charge carrier mobility rather than an increase in charge carrier population. This is significant for radiation detection applications because it indicates there could be a temperature regime where amorphous semiconductors could have significantly enhanced performance. 3) We developed the capability to perform automated, high-precision, electronic property (conductivity, resistivity, Hall mobility) measurements of highly resistive materials as a function of temperature, source voltage, and source current. 4) Three refereed journal papers were published. A fourth paper has been submitted 5) We mentored an undergraduate student and contributed towards their educational and professional development as they assisted in making and testing specimens. 6) The collaborative working relationship was developed with Prof. Angus Rockett at the University of Illinois at Urbana-Champaign (UIUC). They performed key materials characterization experiments to evaluate the electrical performance of different metallization schemes.

The development of Schottky barrier contacts for the higher conductivity amorphous semiconductors was identified as one of the most significant goals to achieve in FY 2008. Initial results looked quite promising, but the results were not reproducible. Significant effort and attention was directed towards this effort by both PNNL and UIUC, but a suitable rectifying contact material and a reproducible deposition process has not been found. Additional details about the status of this task are addressed in the body of this report.

The project was scoped and budgeted to focus on materials synthesis and characterization. The materials and processing challenges associated with developing high-performance contacts on novel semiconducting materials was beyond the scope of this project. That said, the strategy was to leverage traditional semiconductor technologies as a best-effort approach. The concept was to use lower resistivity materials with better charge carrier mobility properties and operate them under reverse bias to create a low-noise, high-resistivity condition that can be switched to a low-resistivity, high conductivity condition

iii

under exposure to radiation events, and then off again. The key to creating effective Schottky barrier contacts lies in finding a metal with a work function that is suitably matched to the semiconductor. Professor Rockett’s research group at the University of IL has been using Kelvin probe force microscopy using an AFM to measure and evaluate the work function of various contact metals and amorphous Cd- Ge-As. They were able to make modest progress on this effort, and their results are presented in the appendix to this report.

Abbreviations and Acronyms

AST Arsenic-selenium-telluride; As40Se60-xTex BSE Back-scattered electron

CVD Chemical vapor deposition

CGA Cadmium-germanium-di-arsenide; CdGexAs2 EDS Energy dispersive spectroscopy

EMSL Environmental Molecular Sciences Laboratory

IR Infrared

IV Current - voltage KPFM Kelvin probe force microscopy NIST National Institute for Standards and Technology NOMSL Non-oxide Materials Synthesis Laboratory PNNL Pacific Northwest National Laboratory PPMS Physical property measurement system SEM Scanning electron microscopy SOW Statement of work UHV Ultra-high vacuum UIUC University of Illinois at Urbana-Champaign UV-VIS-NIR Ultraviolet visible near infrared XRD X-ray diffraction

Z Atomic number

Contents

1.0 Introduction ...... 1.1

2.0 Materials Synthesis ...... 2.1 2.1 Synthesis Method ...... 2.1 2.2 New Chemistries ...... 2.2 2.3 Progress on Identifying the New Metastable Phase ...... 2.4 2.4 Processing Summary ...... 2.5

3.0 Electrical Property Characterization ...... 3.7 3.1 Current-voltage curves ...... 3.7 3.2 Hall mobility ...... 3.8

4.0 Radiation Response Testing ...... 4.8 4.1 DC Ionization Tests ...... 4.9 4.2 Pulse Testing ...... 4.12 4.3 Summary ...... 4.14

5.0 Collaboration with the University of Illinois at Urbana-Champaign (UIUC) ...... 5.1 5.1 Conclusions of the materials characterization at UIUC ...... 5.1 5.2 Schottky contacts ...... 5.1

6.0 Outcomes and Future Direction ...... 6.1 6.1 Major Project Outcomes ...... 6.1 6.2 Future Direction ...... 6.1

7.0 References ...... 7.1

8.0 Appendix A ...... 8.2

vii

Figures

Figure 1. Top: double containment schematic; Bottom: double containment ampoule with copper...... 2.2

Figure 2. Plot of density and conductivity as a function of Te concentration for As40Se(60-X)TeX [7,8]...... 2.3 Figure 3. Microscopic analysis of metastable phases formed in Cd-Ge-As during quenching. Polarized reflected light microscopy (A) revealed two distinct crystal morphologies (polycrystals and needles) in an amorphous matrix. Examination by XRD detected

CdGeAs2 and a new, unidentified phase(s). Electron backscattered diffraction (B) determined the polycrystals (colored) were CdGeAs2, and confirmed the matrix was amorphous, but diffraction patterns from needles were unidentifiable. Compositional

differences (Zave contrast) were accentuated using SEM BSE (C), and stoichiometry was determined by quantitative EDS mapping (D, polycrystals = CdGeAs2, needles = Cd3Ge2As4). Lattice parameters, space group, and structure factor calculations for the new phase are in progress...... 2.4 Figure 4. Electrical conductivity as a function of temperature for chalcogenide amorphous semiconductors. Note the greatly increased conductivity at ~ -20°C...... 3.8

Figure 5. Current-voltage curve for CdGe0.85As2 from 0-100 V at – 40°C ...... 4.9

Figure 6. Current vs. time for As40Se48Te12 as a function of exposure to a sealed source at room temperature and a bias of 500V...... 4.10 Figure 7. Variation in DC ionization current as a function of applied field...... 4.11

Figure 8. Detector gain as a function of applied field for amorphous As2Se3 for exposure to a sealed alpha source...... 4.12 Figure 9. Pulse measured by As-Se-Te detector from an 241Am source. The data was collected using a digital oscilloscope with the detectr biased at 700V...... 4.13 Figure 10. Pulse height spectrum from an 241Am alpha source as collected by an amorphous semiconductor...... 4.13 Figure 11. A test pattern and contact arrays of Mg metal deposited on Sample ASGRAD 41c. Current - voltage measurements were made from Mg contact to Mg contact, from Mg contact to an ohmic contact (used for Hall effect) on the back side of the wafer, and from Mg contact to a probe placed directly on an exposed area of the CGA...... 5.2 Figure 12. Left, a linear scale plot and right a logarithmic plot of the current/voltage curves for sample 41c measured between two sets of Mg contacts in the light and in the dark. A large difference in conductivity is observed...... 5.2

viii

Tables

Table 2.1. Matrix of Iodine-doped Aresenic Trisulfide Compositions Synthesized and Tested.Table ...... 2.3 Table 2.2. Summary of ampoules produced in FY08 ...... 2.5

1.0 Introduction

The guiding motivation of this project was to develop new, innovative materials to enable detection of gamma radiation. The two principal techniques used to detect and obtain energy resolution information from gamma radiation are based on either scintillators or semiconductors. This project chose to focus on the second – the direct conversion of gamma photons into an electrical signal using semiconductors. Traditional semiconductor-based detectors rely on single crystals as the active detecting medium because they have a well-ordered electronic structure that facilitates rapid charge carrier transport. However, there are significant processing problems associated with the growth of large (~ 1cm3), defect-free, single crystals from multi-component materials (e.g., Cd-Zn-Te (CZT) crystals). These challenges limit the size and availability of the crystals for applications. A significant amount of research and development is still required to improve the performance and production yield of single crystal CZT. In light of this need, we have proposed the development of amorphous semiconductors as a potential alternative or interim solution for room-temperature gamma radiation detection. Compared to single crystals, amorphous semiconductors have the potential advantages of rapid, cost-effective, bulk- fabrication; near-net-shape fabrication of complicated geometries; compositional flexibility, and greater compositional and electronic property control. The main disadvantage of an amorphous semiconductor is reduced-charge carrier mobility due to the disordered structure. In this project, we are focusing our efforts to mitigate the technical problems with amorphous semiconductors to develop an optimized material for radiation detection. The materials selected for this study were based on the following criteria: semiconducting, moderate-to- high atomic number, moderate band-gap, high-resistivity, and glass forming. This lead us to two families of amorphous semiconductors – the chalcogenides (compounds based on S, Se, or Te) and the chalcopyrites (a class name given to a family of component compounds that have a tetragonal unit cell)

Within the chalcopyrite family of semiconductors, CdGeAs2 is the best-known glass-former. Single 4 2 crystal CdGeAs2 has very excellent charge carrier mobility ( ~ 10 cm /(Vs) at room temperature) [1,2], and has one of the highest reported non-linear optical coefficients. It is also iso-structural with another multi-component single crystal chalcopyrite semiconductor, AgGaSe2, which has demonstrated radiation detection capability [3]. In the amorphous form, CdGeAs2 has a large reported compositional range (Ge = 0.2 - 1.3) and also has reported substitutional ability (e.g. replace Ge with other elements such as Si or Sb) [4-6]. These properties make it a very good candidate for investigating the potential application of amorphous semiconductors for radiation detection applications. Unfortunately, its band gap (~ 0.9 – 1.0 eV) and its resistivity (107 – 109 Ohm-cm) are somewhat lower than optimal. However, because it is a known multi-component, glass-forming semiconductor, it lies within the targeted region of the periodic table, and it has very large compositional flexibility, it is a good material to demonstrate compositional control of gamma radiation detection properties. Additionally, the fact that it is iso-structural with other higher Z chalcopyrites such as AgGaSe2 (that has been reported to respond in crystal form to gamma radiation) may allow us to blend these compounds together in an amorphous state to create an ideal amorphous gamma radiation detection material. The chalcogenide family of amorphous semiconductors from the As-Se-Te system was chosen not only because of their electrical properties, but because they are excellent glass formers, they are exceptionally resistive (≥1012 ohm•cm), and they are sensitive to optical photons (photo-induced property changes). It was suspected that their photosensitivity might be an indication of being sensitive to ionizing radiation as well (gamma photons).

1.1

The major successes in FY06 were processing based: we developed the ability to synthesize crack-free, amorphous ingots of Cd-Ge-As glass with very high purity and homogeneity. Two different process methods were developed, and various physical, chemical, and electrical analyses were completed to establish the homogeneity and chemical purity of the material. Initial testing and evaluation these material established they had suitable properties for radiation detection applications. In FY07, we were able to move beyond transforming raw elements to new materials, into actually making and testing devices made from these new materials. The major successes in FY07 are as follows: • We demonstrated the ability to measure a DC ionization response to alpha radiation in three different amorphous semiconductors: CdGe0.85As2, As40Se60, and As40Se48Te12. • We demonstrated the ability to control and reduce the density of defect states in the band gap of Cd-Ge-As glasses by controlling composition as well as processing conditions. • We demonstrated the ability to reduce the density of defect states by doping the material with hydrogen, a commonly used approach when working with amorphous silicon. • We were able to build Schottky barrier contacts on Cd-Ge-As glass and demonstrated diode performance and enhanced photosensitivity. The barrier contacts reduced the leakage currents by approximately 5 orders of magnitude. This enabled the ability to measure a photoconductive response in the material: the illuminated conductivity was about 2 orders of magnitude greater than in the dark. In FY08, our major accomplishments included the following: • We continued to refine and improve our materials processing technologies for synthesizing bulk amorphous semiconductors using the double-containment quench method. • We conducted high precision electrical characterization experiments on our specimens and discovered an anomalous temperature region of increased conductivity for As-Se, and As-Se-Te amorphous semiconductors. • We designed and built more sophisticated test devices and fabricated more sophisticated test specimens to include such features as guard rings and pixilated contacts for enhanced radiation detector performance. We performed basic electrical characterization of our test specimens to determine their performance parameters (bias voltage, time constant, etc.) for radiation testing, and then perform radiation tests on them such as DC ionization tests. • We continued to collaborate with UIUC to develop Schottky barrier contacts and build FET structures on Cd-Ge-As. Initial electrical testing results were promising; the devices showed a high degree of photosensitivity, but the results were not reproducible with other specimens. • We conducted successful pulse testing with As-Se, and As-Se-Te amorphous specimens and were able to collect individual pulses as well as pulse height spectrum that showed response to irradiation with alpha sources. The results from this project have increased our understanding of chalcopyrite and chalcogenide amorphous semiconductors. We have shared this increased knowledge with the technical community via published journal articles. This information will contribute to the eventual design of suitable devices and identify suitable applications for using amorphous semiconductor-based radiation detectors.

1.2

2.0 Materials Synthesis

One of the key milestones of success in the project was the ability to form bulk, crack-free amorphous semiconductor ingots. Crystallization is a natural process that occurs as a liquid cools to a solid, and the formation of crystals is driven by changes in free energy of the system: the crystallized state has lower free energy than a disordered solid state. Since changes in free energy favor crystallization, special processing techniques are required to solidify a liquid into a disordered solid state without crystals. In essence, the goal is to rapidly cool the liquid to “freeze” in the disordered liquid structure before the system is able to order and form crystals. We were able to develop a couple of different methods to accomplish this, depending on the crystallization kinetics of the specific material being processed.

2.1 Synthesis Method One approach to fabricating amorphous, crack-free ampoules was to use a double-containment system. Two concentric, evacuated ampoules are used; the inner is filled with the elements of choice, and the gap between the inner and outer is filled with a high thermal conductivity material (e.g. Cu). The approach has several advantages. The increased thermal mass of the system reduces the heat loss during transfer of the ampoule from the furnace to the quench bath. It is possible for the onset of nucleation to occur quite rapidly, so it is important to make the temperature change from furnace to quench bath as steep as possible. For the quench process, the Cu filler provides high thermal conductivity and rapid, uniform cooling throughout the ingot. Additionally, the Cu powder sinters during thermal processing, and provides an additional constraint on the ingot that inhibits volumetric expansion upon crystallization (useful if the amorphous phase is more dense than the crystalline). A cartoon of the design is shown in Figure 1. This technique was required to quench the CGA materials into an amorphous solid state.

Another, more direct approach to form an amorphous solid is to use a single-wall ampoule and to quickly remove the ampoule from the furnace while the liquid is molten and rapidly cool it. Some materials, with a slow crystallization rate, such as silicate glasses, can be slowly cooled in air and the material will still form an amorphous solid. Other materials need to be rapidly cooled by immersing them in a liquid to quench them. The arsenic-selenium-tellurium (AST) materials had slow enough crystallization kinetics that they could be quenched into a disordered solid by cooling or quenching them in air.

2.1

Figure 1. Top: double containment schematic; Bottom: double containment ampoule with copper.

2.2 New Chemistries The Cd-Ge-As chemical system was the primary material family studied during the course of this project. In addition to CGA glass, other amorphous semiconductors were made from the chalcogenide family. There are three primary reasons for choosing this material system. First, chalcogenide glasses are known photosensitive materials, and have many properties that can be manipulated via exposure to sub-band-gap illumination. We project that this sensitivity to optical photons may also extend out to higher photon energies such as gamma or x-rays. Second, amorphous chalcogenide semiconductors also have physical properties that match the previously stated boundary conditions for a good gamma detector (moderate to high Z, moderate band gap, and high resistivity). And thirdly, from a processing point of view, chalcogenide glasses are good glass formers, and can be easily quenched into a disordered solid state.

Three compounds in particular were chosen, As2Se3, iodine doped As2Se3, and As40Se48Te12. Arsenic trisulfide was studied as a base-line chalcogenide material, because there has been extensive optical and electrical characterization of this material, and because is also has a well-know photoelectrical properties (e.g. it is used in photo copiers). Tellurium addition to arsenic trisulfide was chosen so as to simultaneously optimize density and resistivity Figure 2. The specific composition of the ternary compound was selected based an extensive survey of publish properties [7-9].

2.2

Figure 2. Plot of density and conductivity as a function of Te concentration for As40Se(60-X)TeX [7,8].

Doping As2Se3 with iodine was motivated by literature reports that stated the addition of iodine improved electrical conductivity forty-fold. Arsenic trisulfide showed a measureable response to ionizing radiation in FY07, but its high resistivity makes it a difficult material to work with, electrically. Hence, the thought was that if the material could be doped with iodine to reduce its resistivity, then it could be tailored into a more useful radiation detector. In addition to doping with iodine, the glasses were also doped with hydrogen. This was done during making the ampoule by evacuating it and back-filling it with an Ar-H2 blended gas. Hydrogen addition was used based on the improvement it made to the optical absorption coefficient spectra for CGA glass.

This task was accomplished using the help of an undergraduate student from Clemson University. A matrix of 5 different compositions were synthesized, and their electrical properties were characterized, Table 2..

Table 2.1. Matrix of Iodine-doped Arsenic Trisulfide Compositions Synthesized and Tested. mol % Iodine Composition Intended Actual

As2Se3 w/o H2 0 0

As2Se3 + H2 0 0

As2Se3 + 0.5 mol% I + H2 0.50 0.45

As2Se3 + 1 mol% I + H2 1.00 0.86

As2Se3 + 5 mol% I + H2 5.00 4.67

It was possible to produce amorphous As2Se3 doped with hydrogen and up to 5 mol% iodine. The effect of H2 addition was a decreased density, change in optical absorption (blue-shifted), and broadening of the XRD spectra. The effect of iodine addition was mostly observed by changes in optical absorption spectra at IR wavelengths. At 5 mol% I, not only was there an IR shifted absorption edge but a variety of other additional absorption bands. Electrically, iodine addition increased the conductivity of the material up to 2 orders of magnitude. A preliminary report written by the summer student is included in Appendix A.

2.3

2.3 Progress on Identifying the New Metastable Phase In FY07, we reported the discovery of a new, metastable Cd-Ge-As compound whose crystal structure had not been reported in the literature. This new compound can be made during double containment synthesis of CGA glass with a composition of CdGe1.0As2.0. A collage of various micrographs showing the new phase is shown in Figure 3.

Figure 3. Microscopic analysis of metastable phases formed in Cd-Ge-As during quenching. Polarized reflected light microscopy (A) revealed two distinct crystal morphologies (polycrystals and needles) in an amorphous matrix.

Examination by XRD detected CdGeAs2 and a new, unidentified phase(s). Electron backscattered diffraction (B) determined the polycrystals (colored) were CdGeAs2, and confirmed the matrix was amorphous, but diffraction patterns from needles were unidentifiable. Compositional differences (Zave contrast) were accentuated using SEM BSE (C), and stoichiometry was determined by quantitative EDS mapping (D, polycrystals = CdGeAs2, needles = 1 Cd3Ge2As4). Lattice parameters, space group, and structure factor calculations for the new phase are in progress .

Reitveldt analysis (a method of determining the crystal structure of a material using powder XRD data) of the unknown metastable phase was attempted by UIUC. However, it was not possible to use this approach, because the technique requires that the exact chemical composition and the unit cell for the crystal structure to be known as the starting point for the analysis. Consequently, we were able to obtain the assistance of a crystallographer at PNNL who was able to fracture a glass ceramic CGA sample

1 Winner of Best Poster Competition at the Materials Science & Technology (MS&T2007), Detroit, MI, October, 2007.

2.4

containing the metastable compound, isolate a piece containing only a couple of small crystals, and perform single crystal x-ray diffraction analysis on it.

2.4 Processing Summary During FY08, synthesis efforts focused on making CGA glasses with hydrogen doping via the double- containment method and water quenching. Compositions with Ge contents of 0.45, 0.65, and 0.85 were able to be made fairly easily. Multiple attempts were made to synthesize CGA glass with Ge = 1.0, however the ingots typically cracked upon quenching, and often contained the needle-like crystals of the metastable phase. The liquid Ga quench method was somewhat more effective at making crack-free, crystal-free ingots of this composition, but the focus was on using the double-containment method based on the improved optical absorption properties associated with this synthesis method, so liquid Ga quenching was not pursued.

Chalcogenide glasses, due to their stronger glass-forming tendencies, were synthesized using single- walled ampoules and water quenching. A variety of different compositions were made containing Te and iodine.

The list of specimens synthesized during FY 2008 is shown in Table 2.2.

Table 2.2. Summary of ampoules produced in FY08

ID Composition

ASGRAD-60 (Cd0.885Zn0.115)- (Ge0.225Si0.225)-As2.00

ASGRAD-61 Cd1.00Ge1.00As2.00

ASGRAD-62 Cd1.00Ge1.00As2.00

ASGRAD-63 Cd1.00Ge0.85As2.00

ASGRAD-64 Cd1.00Ge0.85As2.00

ASGRAD-65 Cd1.00Ge1.00As2.00

ASGRAD-66 As40Se60

ASGRAD-67 As40Se48Te12

ASGRAD-68 As40Se48Te12

ASGRAD-69 Cd1.00Ge1.00As2.00

ASGRAD-70 As40Se60

ASGRAD-71 Cd0.885Zn0.115Ge0.225As2.00

ASGRAD-72 Cd0.885Zn0.115Ge0.225Si0.225As2.00

ASGRAD-73 Cd1.00Ge0.45As2.00

ASGRAD-74 Cd1.00Ge0.65As2.00

ASGRAD-75 As40Se60

ASGRAD-76 Cd1.00Ge0.45As2.00

ASGRAD-77 Cd1.00Ge1.00As2.00

ASGRAD-78 As40Se60

ASGRAD-79 As2Se3 + 0.5 mol% I + H2

ASGRAD-80 Cd1.00Ge0.85As2.00

2.5

ID Composition ASGRAD-81 As2Se3 + 1.0 mol% I + H2

ASGRAD-82 Cd0.885Zn0.115Ge0.225Si0.225As2.00

ASGRAD-83 As2Se3 + 5.0 mol% I + H2

ASGRAD-84 As2Se3 + H2

2.6

3.0 Electrical Property Characterization

Semiconductor based radiation detectors require high resistivity to minimize noise and allow the detection of small signals from a single gamma event, yet they also need reasonably high charge carrier mobility so that pulses of charge carriers can be collected from discrete gamma events without overlapping each other. Thus, we characterized the electrical properties of the amorphous semiconductors we synthesized in order to evaluate them, and determine which would be the most suitable for developing into a radiation detector.

3.1 Current-voltage curves Measuring the relationship between current and voltage as a function of temperature is one of the more basic and essential semiconductor characterization tests. The data obtained provides information about the conductivity of the material, and the activation energy for conduction. This test was done by creating a linear array of four contacts on the specimen, and by using a high-precision source meter (Keithley 6430) to create a constant flow of current between the outer two contacts, and measuring the voltage drop across the two inner contacts. The temperature would then be changed, and the system allowed to equilibrate and the voltage drop for a constant current flow would be measured again. Though a common test, working with high resistivity materials (> 109 ohm-cm) requires the use of high precision equipment in order to measure and generate the extremely small, trace currents (pA) and voltage signals involved. Another complicating factor is that the slow mobilities characteristic of amorphous semiconductors resulted in long transients when the source current or temperature was changed.

One of the more important observations from our electrical characterization studies was the increase in electrical conductivity as the specimen temperature was decreased to approximately -20°C. The expected response is a continuous decrease in electrical conduction with a reduction in temperature. A rigorous physics-based explanation for this anomalous increase in conductivity has not yet been developed. However, this behavior is very intriguing, and noteworthy. Other researchers have reported similar phenomena in other chalcogenide systems. [10,11]

3.7

Figure 4. Electrical conductivity as a function of temperature for chalcogenide amorphous semiconductors. Note the greatly increased conductivity at ~ -20°C.

3.2 Hall mobility Hall mobility is a technique used to measure the mobility of charge carriers. An array of four contacts is applied to the specimen in a diamond pattern. The specimen is placed in a magnetic field with the line of flux passing perpendicular to the plane of the sample. A fixed current is sourced to flow between two of the contacts (e.g. left to right) while the voltage is measured across the other two contacts (e.g. top to bottom). As the charge carriers flow across the sample, the applied magnetic field causes them to deflect, creating an induced voltage that can be measured. This is called the Hall effect. The test is done in a variety of permutations (e.g. sourcing current left to right, right to left, magnetic field pointing up, magnetic field pointing down, etc.). The result of the data analysis is that the Hall mobility of the charge carriers in the sample can be determined. This is one way to determine the mobility of the charge carriers.

4.0 Radiation Response Testing

Alpha particles provide a convenient way to characterize the performance of a detector for several different reasons. First, they provide a 100% probability of interaction; second, each particle deposits 100% of its energy, enabling an accurate calculation of total flux; third, the location where the energy was deposited can be known, thus enabling computation of charge transport properties; and fourth, the influence of the radiation can be easily controlled or turned “on and off” using a simple shutter. Because of these experimental testing advantages, sealed alpha sources were chosen as a starting point for materials characterization.

4.8

4.1 DC Ionization Tests The first set of experiments performed involved measuring the current voltage characteristics of a specimen with and without exposure to an alpha source. Several different experiments were done with three different materials, one was a Cd-Ge-As glass, and the other two were As-Se-Te glasses. Figure 5 shows the current-voltage curve for CdGe0.85As2 sample at – 40°C over 0-100V. The change in current vs. voltage curve with and without exposure to the sealed source was pronounced – an increase of 36.4 nA.

Figure 5. Current-voltage curve for CdGe0.85As2 from 0-100 V at – 40°C The next set of experiments involved measuring the transient response for exposure to alpha radiation. Direct current ionization tests were done on two different chalcogenide amorphous semiconductors at room temperature. The specimens were slowly biased up to a high voltage (≥ 500V), and a sealed alpha source (secured to a pivoting platform above the sample) was rotated over the top of and alternately away from the specimen. When the sealed source was over the specimen, there was a rapid increase in the measured current, almost instantaneously, that essentially reached a plateau value. After waiting for a couple of minutes, the sealed source was rotated away from the specimen, and the current again almost instantaneously decreased back to its base-line value. This test was repeated several times, thus forming a square wave function. A plot of the data for As40Se48Te12 is shown in Figure 6. The data shows that the high bias applied to the sample was able to fill most if not all of the trap states such that when exposed to ionizing radiation, the sample became significantly more conductive, as the increase in ionization induced electron hole pairs traversed across the sample. When the ionizing radiation was removed, the population of electron hole pairs was reduced, and the conductivity rapidly returned to its base line value. The three cycles show practically no drift in the base line current level or any evidence of hysteresis. Evidence of incomplete trap state filling would have resulted in a waveform that was not square, but rather with a

4.9

curved rise time and a curved decay. The ability to fully populate trap states is an important step in improving charge carrier mobility.

Figure 6. Current vs. time for As40Se48Te12 as a function of exposure to a sealed source at room temperature and a bias of 500V.

Additional experiments were done to evaluate the variation in DC ionization current as a function of applied field. These tests were done with an amorphous specimen of As2Se3. The results in Figure 7 show that as the field increased, the DC ionization current also increased. This response holds promise that improved performance with this material may be possible by using higher biases.

4.10

Figure 7. Variation in DC ionization current as a function of applied field.

Because the specimen was exposed to alpha radiation from a source with known activity and the sample testing geometry was known, the amount of energy deposited into the specimen could be calculated by taking into account various attenuation factors. The energy coming out of the sample could be directly computed from the measured DC ionization current (ΔI due to exposure to the source) and the applied bias voltage. A ratio of the energy output divided by the energy deposited was then computed, and the results are shown in Figure 8. This data clearly demonstrates that a signal gain or amplification can be achieved in this material. Thus, critical charge carrier losses due to trapping phenomena can be compensated for by applying a high enough gain (~ 7000 V/cm), thereby improving charge carrier collection performance. Additionally, by applying an even higher bias voltage, the material can be used as a solid state photomultiplier, and can be tuned to create a gain in signal strength greater than the energy deposited by the incident radiation. As such, these materials maybe suitable for use in applications as photodiodes coupled to scintillator detectors. These same results were reproduced with As40Se60 glass for different configurations of contacts and electrodes.

4.11

Figure 8. Detector gain as a function of applied field for amorphous As2Se3 for exposure to a sealed alpha source.

4.2 Pulse Testing In addition to performing DC ionization experiments, pulse height testing was also done with selected specimens. Individual pulses (Figure 9), as well as time-lapse pulse height spectra were collected (Figure 10). Note that due to limitations in the amplifier electronics (the upper limit of the time constant was too short), the full pulse shape was truncated. Consequently, the amplifier did not collect the full height of the individual pulses as shown in Fig 1, but only a portion of the leading edge. Updated amplifiers should be able to collect the full pulse, thus producing a superior pulse height spectrum. In both instances, very distinct pulses were measured and characterized. This was one of the most significant achievements of the project.

4.12

Figure 9. Pulse measured by As-Se-Te detector from an 241Am source. The data was collected using a digital oscilloscope with the detector biased at 700V.

Figure 10. Pulse height spectrum from an 241Am alpha source as collected by an amorphous semiconductor.

4.13

4.3 Summary Overall, we have demonstrated that measuring DC ionization conductivity is a viable experimental method of screening of glasses for their potential application as gamma detectors. By cooling Cd-Ge-As glass, we were able to demonstrate measurable radiation sensitivity by comparing I-V curves collected with and without exposure to alpha radiation. The room temperature transient response to radiation was studied by measuring the DC ionization current with As40Se48Te12. The sharp step changes between source on and off conditions indicates that the high bias may be sufficient to fill trap states such that ionized electron hole pairs are able to move across the material with higher mobility. The effect of varying applied field on DC ionization current measurements was studied using As40S60. The data demonstrates that not only is it possible to compensate for charge trapping losses, but that actual signal gain can be attained by operating at high biases. Actual direct pulses, and pulse height spectra were collected for As40Se48Te12, which was one of the more significant results from the project. Thus, in spite of the short-range order and trapping issues, these amorphous semiconductors show a measurable response upon exposure to ionizing radiation, and consequently show promise for application as radiation detector materials for either direct detection of radiation or as photodiodes coupled to scintillators.

4.14

5.0 Collaboration with the University of Illinois at Urbana-Champaign (UIUC)

Collaboration with the University of Illinois at Urbana-Champaign (UIUC) focused on contact development (materials selection, deposition strategy, photo-lithography, etc.). Modest gains were made. The results of performing Kelvin Probe Force Microscopy (KFPM) to characterize the work function of CGA specimens are presented as a journal article in the appendix.

5.1 Conclusions of the materials characterization at UIUC Based on materials analyses described previously, we conclude that the material is a typical amorphous semiconductor with a mobility gap in the range of 0.6-0.9 eV, a very broad band edge with a relatively low density of states near the mobility gap. The XRD data shows that the material has a wide range of interatomic distances. Although the structural and chemical properties are very uniform, the electrical and optical properties are not. The mobility gap and resistivity vary from sample to sample based on processing and composition variations. Difficulties in obtaining good quality electrical contacts have also introduced an additional complication. This may indicate variability in the nature of the material on the microscale sufficient to result in large variations in optical and electronic properties.

Amorphous hydrogenated Si (a-Si:H) is a typical amorphous semiconductor. The hydrogen reduces the density of states near the middle of the gap, making it much more straightforward to dope and making it a better semiconductor. Both a-Si:H and CGA glass are tetrahedrally bonded amorphous semiconductors. Thus, using a-Si:H as a model, a specimen of CGA glass was prepared with a hydrogen atmosphere in the process ampoule. The results show that when compared to a specimen of the same composition and under identical conditions, that a reduction in defect density and improved optical properties was obtained.

In general the conductivity of the CGA samples tested to date has been too high for room temperature radiation detection – cooling to -40°C was required to sufficiently reduce the leakage current. Several approaches to improving the situation are possible. One could investigate other processing and compositional changes that could have a greater impact on reducing the density of mid-gap states, similar to the hydrogen experiment described above. An additional approach would be to make diode-like (Schottky) contacts to the material. The depletion regions associated with these contacts can reduce local conductivity and provide a field to collect photocarriers generated by gamma rays.

5.2 Schottky contacts We have begun investigation of the possibility of reducing the sample conductivity through contact junctions. We previously described how high work function metals such as Ti, Au, and Ag typically produce ohmic contacts to the CGA samples. Therefore the best potential for a high barrier Schottky contact would be to use a low work function metal. Typical examples used in organic electronic devices are Mg or Ca.

5.1

We have deposited and patterned Mg contacts on CGA samples as shown in Figure 11. The contacts were then studied by current/voltage measurements from Mg to Mg, from Mg to a Ag ohmic contact used for Hall effect measurements on the opposite side, and from Mg to a probe resting directly on the CGA.

Figure 11. A test pattern and contact arrays of Mg metal deposited on Sample ASGRAD 41c. Current - voltage measurements were made from Mg contact to Mg contact, from Mg contact to an ohmic contact (used for Hall effect) on the back side of the wafer, and from Mg contact to a probe placed directly on an exposed area of the CGA.

The most dramatic results in measurements so far are a large photoconductive response between adjacent Schottky barrier contacts (Mg to Mg) when the sample is exposed to white light. An example of this behavior is shown in Figure 12. Upon exposure to white light, the current increased by over 2 orders of magnitude!

Figure 12. Left, a linear scale plot and right a logarithmic plot of the current/voltage curves for sample 41c measured between two sets of Mg contacts in the light and in the dark. A large difference in conductivity is observed.

Because transient photoconductivity is precisely the expected result of a gamma-ray detection event, this result is very exciting as a possible gamma ray detector.

5.2

The major problem with the Schottky contact study to date is the large variation in device behavior from contact to contact and the relative delicacy of the Mg contacts. The latter is easily solved by over-coating the Mg with another protective metal such as Al and/or Au. The former is more troublesome. However, if these issues can be resolved, it should be possible to build a high-performance radiation detection device.

The strategy for follow up on the preliminary measurements is to obtain new double-polished CGA samples and to deposit Mg or Ca contacts in small patterns on both sides of the sample. Measurements will be conducted both across the sample surface from contact to contact and through the thickness of the CGA. If a reproducible contact pair showing good photoconductivity through the thickness of the sample can be obtained, these will be tested for x-ray and gamma ray sensitivity.

Measurements will be conducted as a function of temperature to determine how the behavior of the contacts changes as the temperature is reduced. This will also supply information about the Schottky barrier height.

5.3

6.0 Outcomes and Future Direction

Based on the progress made during the course of this project, several potential directions for future research in amorphous semiconductor radiation detector research are discussed.

6.1 Major Project Outcomes Four major outcomes resulted from this project: 1. Materials processing techniques were developed for synthesizing bulk amorphous semiconductors from Cd-Ge-As As-Se-I, and As-Se-Te glass were developed. These efforts lead to innovative quenching technologies. A new metastable phase in the Cd-Ge-As family was discovered and reported.

2. Electrical characterization of two amorphous semiconductors indicated an anomalous high conductivity regime at moderately cool temperatures (-20°C). Funds were not available to fully explore this anomaly. Two tentative explanations were offered: 1) The increased conductivity was due to a semiconductor-to-metal transition, or 2) The increased conductivity was due to increased charge carrier mobility due to freezing out of vibrational phonon modes typically responsible for scattering charge carriers at room temperature. The many potential applications that could be developed from materials with these properties warrant additional funded investigation.

3. A protocol, testing apparatus, signal processing equipment, and procedure for conducting DC ionization experiments was developed. Several different amorphous semiconductors (Cd-Ge-As and As-Se/As-Se-Te) were tested with this method and demonstrated measurable, repeatable radiation response using sealed alpha sources.

4. The transient response of two different amorphous semiconductors to radiation at RT was demonstrated using sealed alpha sources. Individual pulses as well as pulse height spectra were collected.

6.2 Future Direction Four main areas of research for continued development of amorphous semiconductor radiation detectors are suggested: 1. Development of Schottky contacts for amorphous semiconductors. This includes the following key steps: a. Identify the correct metal (or sequence of metals e.g. Mg-Al-Au) to use to create a charge-depleted interface, b. Determine the appropriate surface treatment to clean and prepare the semiconductor c. Develop a suitable photolithographic and metallization process to create the contact

2. Sustained electrical characterization of the anomalous high conductivity regime found in As-Se, and As-Se-Te amorphous semiconductors to develop a physics-bases explanation for the phenomena and rule out experimental artifacts.

6.1

3. Development of a thermoelectrically cooled amorphous semiconductor testing apparatus to take advantage of the improved conductivity (charge carrier mobility) at modestly cool temperatures.

4. Continued DC ionization and transient pulse testing of the advanced amorphous semiconductors developed as a consequence of improved Schottky contacts, and testing of thermoelectrically cooled detectors.

6.2

7.0 References

(1) Rud, V.; Rud, Y.; Poluushina, I.; Ushakova, T.; Iida, S., "Observation of Record Electron Hall Mobility in CdGeAs2 Single Crystals," Jpn. J. Appl. Phys, 39 [Suppl. 39-1] 266-267 (2000).

(2) Rud, V. Y.; Rud, Y. V.; Pandey, R.; Ohmer, M. C., "Evidence of high electron mobility in CdGeAs2 single crystals"; pp. 439 in Infrared Applications of Semiconductors III. Symposium (Materials Research Society Symposium Proceedings Vol.607), Edited by Mater. Res. Soc, Boston, MA, USA, 2000.

(3) Roy, U. N.; Groza, M.; Cui, Y.; Burger, A.; Bell, Z. W.; Carpenterl, D. A., "Crystal growth, characterization and fabrication of AgGaSe2 crystals as a novel, material for room-temperature radiation detectors"; pp. 177 in Proceedings of SPIE - The International Society for Optical Engineering, Vol. 5540. Edited by International Society for Optical Engineering, Bellingham, WA 98227-0010, United States, Denver, CO, United States, 2004.

(4) Hruby, A.; Stourac, L., "Semiconducting glasses based on CdAs2," Materials Research Bulletin, 4 [10] 745-756 (1969).

(5) Risbud, S. H., "Processing and properties of some II-IV-V2 amorphous and crystalline semiconductors," Applied Physics A: Materials Science & Processing, 62 [6] 519-523 (1996).

(6) Stourac, L., "Thermal conductivity of semiconducting amorphous CdGeAs2," Czech. J. Phys., V19 [5] 681-684 (1969).

(7) Kokorina, V. F., Glasses for infrared optics. CRC Press, Boca Raton, FL, 1996.

(8) Vengel, T. N.; Kolomiets, B. T., "Vitreous Semiconductors, some properties of materials in the As2se3 - As2Te3 System II," Soviet Physics Technical Physics, 2 2314-2319 (1957).

(9) Mahadevan, S.; Giridhar, A.; Rao, K. J., "Study of electron transport in As-Se-Te glasses," J. Phys. C: Solid State Phys., 10 4499-4510 (1977).

(10) Qamhieh, N.; Willekens, J.; Brinza, M.; Adriaenssens, G. J., "Anomalous DC dark conductivity behaviour in a-Se films," J. Phys.: Condens. Matter, 15 L631-L635 (2003).

(11) Kushwaha, N.; Shukla, R. K.; Kumar, S.; Kumar, A., "Anomalous behaviour in dark conductivity and photoconductivity in a-Se85Te15 − xPbx thin films at low temperatures," Materials Letters, 60 3260- 3264 (2006).

7.1

8.0 Appendix A

Journal articles that were generated as a result of NNSA funding for this project:

(1) Johnson, B. R.; Riley, B. J.; Sundaram, S. K.; Crum, J. V.; Henager, C.; Zhang, Y.; Shuthanandan, V.; Seifert, C. E.; Ginhoven, R. M. V.; Chamberlin, C.; Rockett, A.; Hebert, D.; Aquino, A., "Synthesis and Characterization of Bulk Vitreous Cadmium Germanium Arsenide," Journal of the American Ceramic Society, 92 [6] 1236-1243 (2009).

(2) (1) Johnson, B. R.; Crum, J. V.; Sundaram, S. K.; Ginhoven, R. M. V.; Seifert, C. E.; Riley, B. J.; Ryan, J. V., "DC Ionization Conductivity of Amorphous Semiconductors for Radiation Detection Applications," Transactions in Nuclear Science, 56 [3] 863-868 (2009).

(3) (1) Riley, B. J.; Johnson, B. R.; Crum, J. V.; Thompson, M. R., "Tricadmium Digermanium Tetraarsendie: A New Crystalline Phase Made with a Double-Containment Ampoule Method," Journal of the American Ceramic Society, 95 [7] 2161-2168 (2012).

(4) Damon N. Hebert, Allen J. Hall, Angel R. Aquino, Angus A. Rockett Richard T. Haasch Bradley R. Johnson, Brian J. Riley, “Kelvin probe force microscopy of metal contacts on amorphous cadmium germanium arsenide”, Journal of the American Ceramic Society, submitted, Nov 2012

8.2

J. Am. Ceram. Soc., 92 [6] 1236–1243 (2009) DOI: 10.1111/j.1551-2916.2009.03001.x Journal compilation r 2009 The American Ceramic Society r 2009 No claim to US government works Journal Synthesis and Characterization of Bulk, Vitreous Cadmium Germanium Arsenide

Bradley R. Johnson,w Brian J. Riley, Shanmugavelayutham K. Sundaram, Jarrod V. Crum, Charles H. Henager Jr., Yanwen Zhang, Vaithiyalingam Shutthanandan, Carolyn E. Seifert, Renee M. Van Ginhoven, and Clyde E. Chamberlin Paciﬁc Northwest National Laboratory, Richland, Washington 99354

Angus A. Rockett, Damon N. Hebert, and Angel R. Aquino University of Illinois at Urbana-Champaign, Urbana, Illinois 61801

Cadmium germanium diarsenide glasses were synthesized in ductor with very high room temperature mobility (104 cm2 3 1 10 Á bulk form (B2.4 cm ) using procedures adapted from the liter- (V s)À ). Additionally, it is one of the few chalcopyrite semi- ature. Several issues involved in the fabrication and quenching of conductorsÁ that can be synthesized as an amorphous material in amorphous CdGexAs2 (x 5 0.45, 0.65, 0.85, and 1.00, where x is bulk form. A program of research was initiated to fabricate this the molar ratio of Ge to 1 mol of Cd) are described. An inno- material and evaluate its suitability for radiation detection ap- vative processing route is presented to enable fabrication of plications. This paper is focused on synthesizing bulk amor- high-purity, vitreous, crack-free ingots with sizes up to 10 mm phous CGA compounds, while another has been written diameter, and 30–40 mm long. Specimens from selected ingots describing the radiation response properties of these materials.11 were characterized using thermal analysis, optical microscopy, Glass formation within the CdGexAs2 system has been re- scanning electron microscopy, energy dispersive spectroscopy, ported for x 5 0.02–1.3.12–18 The basic glass forming network particle-induced X-ray emission, Rutherford backscattering, unit for this amorphous compound has been attributed to secondary ion mass spectrometry, X-ray diffraction, density, CdAs2 and glass formation has been shown with the addition and optical spectroscopy. Variations in properties as a function of a number of different ternary elements (Si, Tl, In, Al, Sb, Mg, of processing conditions and composition are described. Results and Ga) as well as Ge.13,15,19 Additionally, substitution of Zn show that the density of defect states in the middle of the band for Cd and Si for Ge has been demonstrated.16 gap and near the band edges can be decreased three ways: The broad compositional flexibility in regard to glass forma- through suitable control of the processing conditions, by doping tion would seem to indicate that this material is a good glass the material with hydrogen, and by increasing the concentration former. Reports in the literature indicate it is possible to air of Ge in the glass. quench bulk samples of CdGeAs2 up to 10 mm in diameter into a vitreous state.17 However, our experience has been contrary. We have found that this is actually a fragile glass-forming sys- I. Introduction tem and that it is very difficult to water quench even modest MORPHOUS semiconductors are the key function material in volumes (10 mm diameter, 30–40 mm tall) of the molten liquid Aseveral important technological areas such as xerography,1 into a bulk vitreous ingot without fracturing it. Although photovoltaics,2 medical imaging plates,3 and TV vidicon tubes.4 quenching molten compounds into a vitreous state is difficult For many applications, thin films are optimal and can be easily enough for fragile glass formers, accomplishing this without fracturing the material because of excessive thermal stresses is fabricated. However for some uses, such as radiation detection, 17 bulk materials are required. Single crystal materials (e.g., Si, Ge, even more challenging. Consequently, it was necessary to de- and CdZnTe) are typically preferred for these applications, but velop new processing techniques in order to produce bulk, there is a need for cost-effective semiconductors that can be op- crack-free, vitreous specimens. The major modification (from erated without cryogenic cooling and that have sufficient bulk typical anoxic material processing procedures that involve evac- volume to efficiently interact with and detect g radiation.5,6 Bulk uated fused quartz ampoules) was the development of a double amorphous semiconductors may be able to contribute to this containment (DC) ampoule method as described below. need, if one can be found with suitable electrical properties. Cadmium germanium diarsenide (CGA or CdGeAs2) belongs to the II–IV–V2 family of chalcopyrite semiconductors. It has II. Experimental Procedures attracted recent attention for its optical properties because of its extraordinarily high nonlinear optical coefficient for second har- (1) Synthesis monic generation (236 pm/V).7–9 However, it is also a semicon- Thermal analysis experiments were used to study the elemental reactions involved with forming CGA compounds and also to B. Dunn—contributing editor measure important glass properties. Reported literature information was used as a starting point to guide experiments.14 Batching reactions of the constituent elements were studied to identify key thermodynamic events. These experiments were Manuscript No. 24976. Received July 12, 2008; approved January 27, 2009. conducted using a Seiko TGA/DTA320 (Northridge, CA), The Environmental Molecular Synthesis Laboratory, a national scientific user facility sponsored by the Department of Energy at PNNL, and the Materials Research Laboratory, which is a combination of thermogravimetric analyzer (TGA) also a national scientific user facility sponsored by the Department of Energy at the Uni- and differential thermal analyzer (DTA). Hermetically sealed versity of Illinois at Urbana-Champaign. This work was supported by the U.S. Department of Energy, Office of Nonproliferation Research and Development (NA-22). Pacific North- aluminum pans were used to minimize material loss because west National Laboratory (PNNL) is operated by Battelle for the Department of Energy of volatilization. A total sample mass between 20 and 30 mg under contract DE-AC06-76RLO 1830. w was used, at a heating rate of 101C/min up to 6201C, and with Author to whom correspondence should be addressed. e-mail: bradley.johnson@ 3 pnl.gov a 150 cm /min purge flow of ultra high purity Ar gas. Glass 1236

8.3

1238 Journal of the American Ceramic Society—Johnson et al. Vol. 92, No. 6 using liquid Ga (701–1501C) as a quenchant. In their work, were performed with a Cameca IMS-5f instrument using an quenching in warm liquid Ga was supposed to provide some O2 ion beam. This analysis condition is highly sensitive to elect- annealing of the sample during the quench and reduce thermal ropositive species and to a broad range of transition and other shock. Although expensive, liquid Ga was an appealing quen- metals. The results were used to corroborate compositional data chant because of its low melting point (29.761C), very high boil- obtained by PIXE and RBS. ing point (22041C), and low vapor pressure.23 DC ampoules Current–voltage curves as a function of temperature were were typically quenched in a water bath at room temperature. measured with a Hall effect testing device consisting of a MMR After quenching, the ingots were annealed at B501C oTg for Technologies analysis head and H-50 van der Pauw controller. that composition (as determined by DTA/TGA/DSC) for B8h The sample temperature was controlled by a heater and a CTI to aid in reducing thermally induced stress in the glasses. After Cryogenics refrigerator with a base temperature of 259 C. The s 1 cooling, the ingots were cast into an acrylic (Buehler VariKleer , sample was mounted on the refrigerator and reheatedÀ with the Lake Bluff, IL) to reduce fracture during sawing. They were heater power controlled by a Lake-shore 330 autotuning tem- then sectioned into thin wafers 1–3 mm thick, polished, and an- perature controller to a set point temperature. A magnetic field alyzed. of 1 T (Varian Associates electromagnet) could be applied to the sample for Hall effect. Data was taken and analyzed using a custom Labview program. Ohmic contacts were made by evap- (2) Characterization orating high work-function metals such as Ti, Au, or Pd onto Specimens were sectioned from the ingots and characterized us- polished CGA disks. Four contacts were applied in a Van der ing a variety of techniques to determine processing-property as Paw geometry. Indium-coated wires were attached to the con- well as composition-property relationships. Optical microscopy tacts using indium solder. Voltages applied to the samples de- was performed with a Leitz Orthoplan optical microscope using pended upon the current to be driven and this current was reflected cross-polarized light, a 1/4 l calcite wave plate, and a minimized based on signal-to-noise requirements as this pro- rotating stage. This enabled the rapid detection of crystalline duced the most reliable results. Typical currents were of the or- phases as the variations in birefringence and crystallographic der of 1 nA for most measurements. The conductivity of the orientation caused the different grains to go in and out of ex- sample was calculated from the I–V data based on the geometry tinction at different angles as the stage was rotated. X-ray of the test. diffraction experiments were conducted with a Scintag PAD V The optical bandgap of these materials were characterized diffractometer using CuKa radiation (l 5 0.15406 nm, 45 kV, using a Varian Cary 5 G UV–Vis–NIR spectrophotometer with and 40 mA) and equipped with a Peltier-cooled Si(Li) solid-state a spectral range of 175–3300 nm and an out-of-plane double detector in a y–2y geometry. Samples were analyzed in a step- Littrow Monochromator. Infrared (FTIR) spectroscopy and scan approach, typically from 101 up to 1101 2y with a step size UV–Vis–NIR spectroscopy were performed on all samples us- of 0.041 2y and a dwell time of 6 s/step. Each sample was ing a Thermo 6700 FTIR and a Varian Cary 500 spectrometer, mounted on a holder that rotated in the x-y plane to minimize respectively. Polished windows of each composition were preferred orientation effects. Samples were analyzed to detect mounted and transmission measurements were collected from and identify the presence of any crystalline phase(s) and for ra- 2.5 to 25 mm and 300 to 2500 nm on the Thermo and Varian dial distribution function analysis. A JEOL 5900 (Joel Ltd., instruments, respectively. Peabody, MA) scanning electron microscope with a Robinson Series 8.6 backscattered electron (BSE) detector and EDAX Genesis energy dispersive spectrometry (EDS) system were used III. Results and Discussion to analyze selected specimens for homogeneity and phase separation. Density measurements were accomplished using a Mi- Thermal analysis studies of the batch reactions between the cromeritics AccuPyc 1330 (Norcross, GA) helium pycnometer. three elements detected three significant endotherms: the melting Composition determination, uniformity, and trace element point of Cd (observed at 3251C) and two additional endotherms contamination experiments were performed using particle- at 5731 and 5921C. Neither of these corresponded to an obvious induced X-ray emission (PIXE) spectroscopy and Rutherford phase transition for one of the elemental constituents. However, backscattering spectroscopy (RBS) on the linear accelerator at the As–Cd phase diagram24 showed several phase transitions the Environmental Molecular Sciences Laboratory (EMSL, associated with the As2Cd3 a, a0 and b phases, with one at Richland, WA). PIXE measurements were carried out with 5781C. Furthermore the CdAs2–As2Cd3 eutectic is given as a 2.5 MeV H1 beam, and the X-rays emitted during the de- melting at 6101C. We hypothesize that the endotherms at 5731 excitation process were analyzed using a Li-drifted Si detector and 5921C represent phase transitions in a Ge-poor As–Cd positioned at an exit angle of 391. The system was calibrated compound or in the ternary CdGexAs2, itself. A distinct, pri- using a standard from the National Institute for Standards and mary, exothermic reaction to form CdGexAs2 was not obvious Technology. The GUPIX computer code, developed at the Uni- in the DTA data but may have been obscured by the rise in the versity of Guelph, Canada, was used to fit the experimental data beginning at the 5731C endotherm and continuing on to the data. Background subtraction to remove bremsstrahlung effects end of the experiment. The use of hermetically sealed-Al pans was performed and the areas under each peak were converted limited the experiments to 6201C. Even at that temperature, re- into concentrations using experimental parameters such as en- actions between the molten Cd, As and/or Ge with the Al pan ergy, incident and exit angles, charge, solid angle of the detector, were noted based on ex situ observations of chemical attack on absorber thickness, and detector response functions. The exper- the pan. Therefore, the observed rising slope in the DTA data imental parameters were then calibrated against the known con- above 5731C probably also included a significant component of centrations of the standard. Following calibrations, PIXE reaction with the Al pan. measurements were collected on CdGexAs2 samples. Subsequent thermal analysis experiments of quenched CGA Rutherford backscattering experiments were performed with glasses of various compositions were able to detect both the 2.0-MeV He1 ions at a scattering angle of 1651. The SIMNRA glass transition temperature and the crystallization tempera- simulations code was used to fit the RBS data. These experi- tures. The results from all of the tests are shown in Table I. ments were performed to determine absolute concentrations of The data from the batch reaction and the glass measurement the major constituent elements. Instead of detecting character- experiments were used to develop the heating rate profile used to istic X-rays, the incident ions that are subsequently scattered process, quench, and anneal the CGA specimens. The results from target atom nuclei are detected and measured. This pro- were similar to the values of a more extensive study by Hong, vides information on the mass and depth of the target atoms. et al.14 Variations between the data sets may be attributed to Additional compositional analyses were conducted using sec- differences in sample volume (smaller with Hong and colleagues) ondary ion mass spectroscopy (SIMS). These measurements and purity (e.g., final ampoule sealing pressure was higher with

8.4

June 2009 Synthesis and Characterization of Bulk, Vitreous CGA 1239

Table I. Glass Transition (Tg) and Crystallization (Tc) Tem- peratures for Various Cadmium Germanium Diarsenide Glasses

Quench rpycn. 3 Name method Composition Tg (1C) Tc (1C) (g/cm )

G 0.45 DC CdGe0.45As2 353 440 5.8094 G 0.65 LG CdGe0.65As2 372.2 463.9 5.7774 G 0.85 DC CdGe0.85As2 373.7 448.5 5.7405 G 1.00 LG CdGe1.00As2 390 448 5.7171

3 5 5 7 Hong’s vs. This study: 10À vs 10À Pa, or 10À vs 10À torr, respectively). Anecdotally, greater success was experienced in obtaining crack-free, crystal-free CdGexAs2 ingots for x 5 0.45 and 0.85 than for the other compositions. In contrast, a greater percentage of ingots for compositions with x 5 0.65 and 1.0 failed due to either cracking or crystallization or both. This was in contrast to the findings of Hong et al.,14 where they computed a maximum in glass formation tendency between x 5 0.3 and 0.6 based on thermal analysis data. Additional thermal analysis experiments at finer compositional intervals are needed to better characterize these variations in glass formability. In addition to thermal glass formability criteria (based on differences between the crystallization and glass transition temperatures) variations in residual oxygen content have also been reported to affect 27 Fig. 1. Polarized reflected optical micrographs of polished cross- glass formability. Hruby et al. stated that when they took the sections of CdGexAs2 ingots. A large-grained, poly-crystalline micro- effort to remove oxygen and water contamination from their structure resulted from air quenching specimens with x 5 1.0 (A). Water starting materials and ampoule walls it was much more difficult quenching the same composition yielded a microstructure with an amor- to form bulk As–Te glasses. Because these Cd–Ge–As glasses phous rim and a smaller-grained polycrystalline interior (B). By reducing were processed at significantly lower oxygen and moisture con- x (0.45–0.85), and using the DC method, amorphous and crack-free in- tamination levels than reported previously, it is possible that a gots were obtained, (C). Similar results were also achieved by quenching similar effect is at work. To our knowledge, the influence of in liquid Ga for values of x up to 1.0 (D). trace oxygen content on glass formability in Cd–Ge–As glass has not yet been rigorously examined. Initial experiments synthesizing bulk CGA glasses were not method, and reducing the Ge content, it was possible to form successful because of failed and cracked ampoules. The source crack-free, amorphous ingots (Fig. 1(C)). The improvement in of the cracking was attributed to Cd attack of the fused quartz glass formability by reducing Ge content was anticipated, based wall. Pyrolytic graphite coatings (described previously) were at- on reports by Hong and colleagues.14,28 Quenching in liquid Ga tempted as a means to address this problem, but yielded mixed was also effective, and an example of a Ge 5 1.0 ingot is shown results. They were effective in eliminating the problem of amin Fig. 1(D). poule failures because of Cd attack of the fused quartz, but they Extensive materials characterization was necessary in order to introduced several additional problems. We observed that the validate that the processing methodology was not introducing CGA melts had a tendency to wet and adhere to the walls of contamination or producing artifacts that would obscure or in- the pyrolytically coated ampoules. This reduced ingot yield, be- terfere with analyzing performance data. Stoichiometric control cause a film of the CGA glass remained along the ampoule wall of the batching and synthesis of CGA glass was validated using rather than pool and collect as a bulk ingot. Additionally, the RBS and PIXE in a complimentary fashion. The absolute ele- pyrolytic coatings themselves were not sufficiently adherent mental concentration of Cd was measured with RBS (Fig. 2). to the ampoule wall, but became incorporated into the melt. Unfortunately, overlap of the data for As and Ge interfered with SEM examination of polished cross sections from ingots made precise measurement of their individual concentrations. Conse- in pyrolytically coated ampoules showed evidence of carbon quently, PIXE was used to identify the elemental ratio of As:Ge contamination in the ingot. Electrical characterization tests per- while normalizing to the RBS measurement for Cd. This method formed by measuring the conductivity as a function of temper- was used to provide accurate compositional measurements ature showed that an ingot synthesized in a pyrolytically coated (Fig. 2). Specimen homogeneity was evaluated by translating ampoule was over an order of magnitude more conductive than the specimen and making measurements at B1.2–1.5 mm inter- the ingot made in a bare ampoule and its conductivity also had vals across the diameter of the disk. Different slices taken from significantly lower temperature dependence. Consequently, pyr- the same ingot were also analyzed to establish homogeneity olytical coatings had to be abandoned. As a result, the specific along the axis of the ingot. The experimental results are shown element loading sequence described previously was developed, in Table II and indicate that stoichiometry was maintained and proved to be an effective means to address the problem of within 71 at.%. The extra, unlabeled peaks in the PIXE plot Cd reacting with fused silica. were attributed to pulse pile-up and escape peak artifacts, based Development of an effective quenching process to form bulk, on complementary analysis of these specimens by SIMS. amorphous, crack-free ingots proved to be challenging. Optical SIMS was chosen for the purpose of corroborating the RBS microscopy (Fig. 1) and XRD (see later discussion) were the and PIXIE results as well as to analyze specimens for trace primary analytical techniques used to evaluate different quench- elemental contamination. Four different specimens were exam- ing methods. Polished cross-sectioned specimens were imaged ined. There was no evidence of any transition metals. Transition using cross-polarized reflected light microscopy (Figs. 1(A) and metal contaminants have been shown to have a measurable (B)). Air quenching, though reported in the literature to be effect on electrical transport properties in CGA glass.29,30 Con- effective,17 yielded only polycrystalline ingots. Water quenching sequently, this type of contamination could seriously interfere was somewhat more effective, but it was only able to produce with electrical characterization experiments, and prevent unam- ingots that were partially amorphous around the perimeter biguous correlations between processing, composition, and (Fig. 1(B)) or cracked. However, by using the DC ampoule property relationships. This data combined with the results

8.5

1240 Journal of the American Ceramic Society—Johnson et al. Vol. 92, No. 6

Fig. 3. X-ray diffraction spectra of CdGe0.85As2 glass verifying their amorphous structure (A) and simulated pair–distribution function for a CdGexAs2 glass for x 5 0.85 (B). Fig. 2. Graph of RBS (A) and PIXE (B) data for a CdGe0.45As2 glass specimen. results are shown in Fig. 3(B). Detailed measurements of the from RBS and PIXE substantiated that the synthesis process PDF nearest neighbor (NN) peak heights for four CGA glasses, was able to accurately achieve the speciﬁed stoichiometry and and the crystalline reference are shown in Table III. The PDF plots for all of the CGA glasses were very similar with the in- that the process was clean, and essentially free of trace contam- ˚ ination down to ppm levels. dividual peaks being located within 70.1 A of each other. The X-ray diffraction results of a CGA glass with 0.85 Ge content trend is that the NN distances increased slightly from 0.45 to are shown in Fig. 3(A), and are typical of all compositions. 0.65 Ge for four measurable NN peaks. However, at 0.85 Ge, Short range order in the CGA glasses was evaluated by calcu- the NN distances decreased below the value for 0.45 Ge for the lating the atomic pair distribution function (PDF) from the ﬁrst 3 NN peaks but increased steadily at the fourth NN peak. amorphous XRD spectra.31–34 For PDF calculations, XRD Overall, the PDF’s show that the amorphous CGA glasses have a high degree of short range order out to the third NN. In com- spectra were collected from 51–1441 2Y at a step interval of parison to the computed PDF plot for the crystalline reference, 0.051 with a dwell time of 3 s/step, and summing up the results of three scans. The spectra were collected using the same diffracto- the G(r) plot for the amorphous CGA was shifted to the right, had shorter NN distances, broader peaks, and lost coherency meter described previously. Atomic PDF calculations were per- ˚ formed using the computer program ‘‘PDFgetX2’’r.34 The after the third NN at 6.3 A. The PDF computed from the crystal structure model of CdGeAs2 consisted of four prominent peaks with a shoulder Table II. Summary of Line Scan Compositional Data on the third one. Peak assignments were based on measurements Computed from Combined RBS and PIXE Analysis for CdGeAs2 Glass (Measurement Accuracy, 70.5%) Table III. Summary of Atomic Pair Distribution Data Spot Cd (at.%) Ge (at.%) As (at.%) for CdGexAs2 0 25.6 24.7 49.6 Ge:Cd ratio First NN Second NN Third NN Fourth NN 1 25.9 24.6 49.5 0.45 2.56 4.09 6.34 7.45 2 25.9 24.6 49.5 0.65 2.57 4.10 6.40 7.49 3 26.0 24.6 49.4 0.85 2.55 4.06 6.32 7.51 4 25.9 24.6 49.4 1.0 2.56 4.10 6.36 7.46 5 25.9 24.7 49.3 w CdGeAs2 2.53 4.35 6.32 /7.23 8.54 6 25.7 24.8 49.5 reference Average: 25.9 24.7 49.5 wShoulder on the third prominent peak. NN, nearest neighbor.

8.6

June 2009 Synthesis and Characterization of Bulk, Vitreous CGA 1241

3 made of the crystal structure model of CdGeAs2 (as published compounds decreases going from 5.86 g/cm for CdAs2 to 5.62 35 3 by Marenkin ) using the computer program ‘‘CrystalMaker’’ g/cm for CdGeAs2 (a pseudo trend-line was drawn as a guide to r. The first peak was located at 2.52 A˚ and was attributed to the eye).12,35,36 A similar variation in density with Ge content the convolution of the As–Ge and As–Cd bonds at 2.425 and was observed for the amorphous specimens fabricated in this 2.633 A˚ , respectively.14,28 The second peak was at 4.35 A˚ study, however, the amorphous compounds were approximately and was assigned to As–As second NN at distances of 4.219 2% more dense than their crystalline counterpart at x 5 1.0. and 4.459 A˚ . Atom pair distances between Cd–As and Ge–As This was consistent with data reported elsewhere.12,14 The shift exist at 6.33 and 6.35 A˚ , thus the shoulder in the crystalline PDF in NN distances to shorter values as observed in the amorphous at 6.32 A˚ was assigned to these atom pairs. The third prominent PDF (Fig. 3(B)) also illustrates this relationship. The variation peak in the crystalline PDF was located at 7.23 A˚ and was as- in density with phase was important to note while interpreting signed to As–As atom pairs, since several As–As atom pair dis- failed experiments, so as to make appropriate processing ad- tances were measured at 7.14, 7.43, and 7.91 A˚ . The fourth peak justments. Density variations with processing method indicated was located at 8.64 A˚ and most likely corresponds to several that the DC-quenched specimens were slightly denser than the different atom pairs (Cd–Cd, Ge–Ge, and As–As) that all occur LG-quenched specimens. Because the amorphous phase was at 8.402 A˚ in the crystal model. denser than the crystalline phase, the higher density may have The first peak of G(r) for the CGA glasses was located at 2.56 been due to the additional constraint imposed by the annular A˚ and was at a slightly farther distance than in the crystal PDF. Cu ring in the DC ampoule, or it may be an indication that the The farther NN distance in the CGA glass is probably due to the DC quench method was slightly faster than the LG method. fact that most of the glasses have a lower Ge content than the Although the macroscopic bulk density decreases with increas- crystal, thus causing a slight peak shift toward the longer As–Cd ing Ge content, it is important to point out that the atomic bond distance of 2.633 A˚ . The second NN peak was located at density (atoms/cm3) increases with Ge content. This is explained approximately 4.09 A˚ for the amorphous G(r), and its position by the change in average atomic mass and in covalent or atomic was 5.6% shorter than the second NN distance for the crystal- radius of the atoms as the alloy changes composition. line specimen. These peaks were assigned to As–As second NN The electronic structure of crystalline semiconductors is mod- distances as was the same for the perfect crystal. The higher eled as having a distinct conduction and valence band edge with relative intensity of this peak is consistent with the fact that these a well-defined band gap. Amorphous semiconductors typically compounds contain at least 50% As. Additional corroborating exhibit bands of extended states with band edges similar to those data are needed to make a definitive assignment of specific atom of the crystalline materials combined with increasingly localized pairs to the PDF peaks in the amorphous CGA glasses beyond states below the band edge whose density decays exponentially the second NN. For example, it is not clear whether the third with energy away from the band edge. The exponentially de- peak in the amorphous PDF at 6.35 A˚ corresponds to the caying density of states is referred to as a band tail. Finally, there shoulder in the crystalline PDF at the same location (e.g., are isolated states in the energy gap (properly the mobility gap), Cd–As and Ge–As atom pairs), or to the prominent third which give rise to a typically constant density of states. The op- peak due to As–As atom pairs located almost an angstrom tical absorption coefficient of the material reflects the product farther away, or possibly the convolution of other possible atom of the density of states in the valence and conduction bands. pairs. Because the glasses contain at least 50% As, and since Therefore, the absorption coefficient typically shows a quadratic the amorphous PDF peak at 4.09 A˚ was shifted to the right, decrease near the energy separating the two sets of extended the third peak in the amorphous PDF may be due to As–As states, an exponentially decaying behavior resulting from the atom pairs, but a peak shift of approximately 0.88 A˚ seems un- band tails, and a constant absorption for energies where only usually large. The fourth peak for the amorphous specimens isolated defect states occur. occurred from 7.4–7.5 A˚ , and was very diffuse. Again, definitive Optical absorption spectroscopy and ellipsometry were used assignment of this peak was not possible. Several As–As atom to characterize the transition from the band edge toward the pair distances in the single crystal structure at 7.14, 7.43, and middle of the mobility gap. At energies higher than the band 7.91 A˚ , so this peak may be due to a convolution of those edge, the material is relatively opaque, and the optical absorp- atom pairs. tion coefficient becomes very large. Therefore a simple trans- The macroscopic bulk densities (g/cm3) of the samples were mission/reflection model will not work. Ellipsometry, however, measured using He pychnometry and were found to vary as a provides a measure of the absorption coefficient out to much function of Ge content, processing method, and structure (i.e., lower transmission values. Toward the middle of the mobility crystal vs. amorphous); see Fig. 4. The density for crystalline gap, the material is more transparent and optical transmission measurements suffice. The greater the density of defect states in the material, the more gradual the transition will be from the mobility edge to the mobility gap, and the higher the absorption coefficient will be in the gap region. Consequently, improved electrical conduction properties depend on minimizing the density of defect states. Ideally the material should be characterized by (1) a steep slope of the absorption coefficient curve at the mobility edge and (2) minimizing the low-energy absorption coefficient. The effects of composition and processing conditions on the density of defect states was thus evaluated. Optical absorption coefficient measurement results for a variety of CGA glasses and processing methods are shown in Fig. 5 and combined optical absorption and ellipsometry results are given in Fig. 6. The effect of composition on the density of defect states was evaluated for both the DC and the LG quench methods. The variation for both processing methods was similar, increasing the Ge content resulted in a reduced optical absorption coefficient (Figs. 5(A) and (B)). There was about a 50% reduction in the absorption coefficient upon changing the Ge content from x 5 0.45–0.85 Ge for the DC process, whereas for the LG process, there was over a factor of 4 reduction in Fig. 4. Variation in bulk density with composition (x), phase, and pro- the absorption coefficient for changing the Ge content from 0.65 cess method for CdGexAs2 glasses. to 1.0.

8.7

1242 Journal of the American Ceramic Society—Johnson et al. Vol. 92, No. 6

Fig. 5. Absorption coefﬁcient data for CGA glasses showing effects of composition (A) and (B), processing method (C) and hydrogen doping (D).

The effect of processing method (DC vs LG) on the density of the maximum value measurable to the constant value far below defect states was also evaluated. Two different Ge compositions the absorption edge. For the 0.85 Ge composition, the absorp- were selected (0.65 and 0.85 Ge), and ingots were made from tion coefficient data were lower by over a factor of two in the each composition using both of the processing methods. Optical gap compared with the highest measurable value. The results absorption coefficients were measured as a function of energy were similar for samples thinned as much as the mechanical for samples taken from each ingot, and the results for 0.85 Ge properties of the material would permit. are shown in Fig. 5(C). For both compositions the results were Reducing the density of defect states in amorphous semicon- the same; the DC quench process yielded a lower optical ab- ductors is a common concern. Amorphous silicon (a-Si) is one sorption coefficient curve than the LG quench process. For a of the most prevalent amorphous semiconductors and is widely composition of 0.65 Ge, the DC process resulted in approxi- used in photovoltaic applications. One strategy used to reduce mately a threefold reduction in the absorption coefficient from the density of defect states in a-Si is to dope it with hydrogen during thin film deposition. The hydrogen acts to passivate dangling bonds. Because CGA glass is a tetrahedrally bonded material just like a-Si, hydrogen doping (as described in the experimental section) was evaluated to determine if it could have a similar positive effect on reducing the density of defect sites. A comparison was made between two different specimens, both having the same composition (0.85 Ge) and both processed in the same method (LG), but one was doped with hydrogen. The optical absorption coefficient measurement results are shown in Fig. 5(D). Hydrogen doping was able to reduce the absorption coefficient values in the midgap energy range by almost a factor of two. The results of the optical absorption data indicates that it is possible to reduce the density of defect states in CGA glass both near the mobility edge as well as in the middle of the mobility gap. This can be achieved by controlling composition, processing conditions, and by doping the material with hydrogen. Hall conductivity measurements as a function of temperature were made on several different CGA glasses. In general, the Hall conductivity trends paralleled those of the absorption coefficent Fig. 6. Combined ellipsometry, optical transmission and reflection data: specimens with lower optical absorption coefficients had data. For the thick samples studied here there is a range of absorption lower Hall conductivity values. Increasing Ge content resulted coefficients that are too low for ellipsometry and too high for absorption in lower conductivity and DC-quenched specimens had lower measurements. conductivity values than LG-quenched specimens with the same

8.8

June 2009 Synthesis and Characterization of Bulk, Vitreous CGA 1243

Ge content. Room temperature conductivity values for DC 5G. F. Knoll, Radiation Detection and Measurement,3rdedition,Wiley,New 7 York, 2000. quenched CGA ranged from 9.1 10À O/cm for 0.45 Ge to 6 7 Â A. Owens and A. Peacock, ‘‘Compound Semiconductor Radiation Detectors,’’ 1.1 10À O/cm for 0.85 Ge. However, hydrogen doping did Nucl. Instrum. Methods Phys. Res. Sect. A, 531 [1–2] 18–37 (2004). notÂ appear to have a significant effect on the electrical transport 7P. G. Schunemann, K. L. Schepler, and P. A. Budni, ‘‘Nonlinear Frequency properties, as there was very little difference in conductivity val- Conversion Performance of AgGaSe2,ZnGeP2,andCdGeAs2,’’ MRS Bull., 23 [7] ues for doped and undoped LG-quenched CGA glasses of the 45–9 (1998). 8E. Tanaka and K. Kato, ‘‘Second-Harmonic and Sum-Frequency Generation same composition. in CdGeAs2,’’ p. 475. Infrared Applications of Semiconductors II. Symposium, In summary, with increasing Ge content, the following trends December 1-4, Boston, MA, 1997. were observed: optical absorption coefficients decreased; Hall 9P. Zapol, R. Pandey, M. Seel, J. M. Recio, and M. C. Ohmer, ‘‘Density Func- conductivity decreased; bulk density decreased; atomic density tional Study of the Structure, Thermodynamics and Electronic Properties of CdGeAs2,’’ J. Phys. Condens. Matter, 11 [23] 4517–26 (1999). increased; and NN distances increased. Hence it appears that 10V. Rud, Y. Rud, I. Poluushina, T. Ushakova, and S. Iida, ‘‘Observation of increasing Ge content modifies CGA glasses by creating a more Record Electron Hall Mobility in CdGeAs2 Single Crystals,’’ Jpn. J. Appl. Phys, 39 [Suppl. 39–1] 266–7 (2000). open structure, making electrical conduction more difficult 11 (assuming a hopping conduction mode) and subsequently de- B. R. Johnson, J. V. Crum, S. K. Sundaram, R. M. V. Ginhoven, C. Seifert, B. J. Riley, and J. V. Ryan, ‘‘DC Ionization Conductivity of Amorphous Semicon- creasing the density of states at the edges and middle of the ductors for Radiation Detection Applications,’’ IEEE Trans. Nucl. Sci., (2009), in conduction band. press. 12L. Cervinka, R. Hosemann, and W. Vogel, ‘‘Relations Between the Structure of Crystalline and Amorphous CdGexAs2,’’ J. Non-Cryst. Solids, 3 [3] 294–304 IV. Conclusions (1970). 13L. Cervinka, A. Hruby, M. Matyas, T. Simecek, J. Skacha, L. Stourac, Two different methods for synthesizing bulk, crack-free ingots J. Tauc, V. Vorlicek, and P. Ho¨ schl, ‘‘The Structure and Electronic Properties of Semiconducting Glasses Based on CdAs2,’’ J. Non-Cryst. Solids, 4, 258–71 (1970). of CGA glass were developed. The use of pyrolytic coatings lin- 14 ing the process ampoules was dismissed due to problems asso- K. S. Hong, Y. Berta, and R. F. Speyer, ‘‘Glass–Crystal Transition in II–IV–V2 Semiconducting Compounds,’’ J. Am. Ceram. Soc., 73 [5] 1351–59 ciated with reduced ingot yield, contamination of the melt by the (1990). 15 coating, and degradation of the resulting electrical properties. A. Hruby and L. Stourac, ‘‘Semiconducting Glasses Based on CdAs2,’’ Mater. Res. Bull., 4 [10] 745–56 (1969). Stoichiometry control, homogeneity, and minimal trace elemen- 16 S. H. Risbud, ‘‘Processing and Properties of Some II–IV–V2 Amorphous and tal contamination of the synthesized ingots validated the ability Crystalline Semiconductors,’’ Appl. Phys. A Mater. Sci. Process., 62 [6] 519–23 of both processes to yield good quality, high-purity amorphous (1996). materials. PDF analysis of CGA glasses showed that they re- 17S. Sharma, K. S. Hong, and R. F. Speyer, ‘‘Glass Formation in Chalcopyrite tained NN atomic coherency out to the third NN. The PDF Structured Semiconducting Compounds,’’ J. Mater. Sci. Lett., 8 [8] 950–4 (1989). 18S. Sharma and R. F. Speyer, ‘‘Optimization of Quench Rate in the Fabrication peaks were broader and located at shorter distances than in the of Bulk Chalcopyrite Glasses in Fused Silica Tubes,’’ J. Mater. Sci. Lett., 8 [3] crystal, but tended to increase in distance with Ge content. 287–90 (1989). Composition-property and processing-property relationships 19A. Hruby, ‘‘The Nature of Glass-Forming Tendency in Cd–As System,’’ were demonstrated using optical absorption spectroscopy and Czech. J. Phys., 26, 593–8 (1976). 20A. Giridhar and S. Mahadevan, ‘‘DC Electrical Conductivity of Some Hall conductivity. The results indicate that the density of defect Cd–Ge–As Glasses,’’ Bull. Mater. Sci., 12 [2] 107–15 (1989). states in the material near the band edges and in the middle 21W. Kern and D. A. Puotinen, ‘‘Cleaning Solutions Based on Hydrogen Per- of the mobility gap can be reduced three different ways: (1) by oxide for Use in Silicon Semiconductors,’’ RCA Rev., 31, 186–206 (1970). 22 increasing the Ge concentration, (2) using the DC quench pro- B. J. Riley, B. R. Johnson, S. K. Sundaram, M. H. Engelhard, R. E. Williford, and J. D. Olmstead, ‘‘Pressure–Temperature Dependence of Nanowire Formation cess, and (3) by doping with hydrogen. The variation in prop- in the Arsenic–Sulfur System,’’ Phys. Chem. Glass, 47 [6] 675–80 (2006). erties with composition was attributed to the effect of Ge on the 23D. R. Lide, CRC Handbook of Chemistry and Physics, Internet Version 2007, short-range order in the material. Thus, we have been able to 87th edition, Taylor and Fancis, Boca Raton, FL, 2007. 24 demonstrate broad control of semiconductor properties for Cd– T. B. Massalski, J. L. Murray, L. H. Bennett, and H. Baker, Binary Alloy Phase Diagrams, Vol. 1. American Society for Metals, Metals Park, OH, 1986. Ge–As glasses by manipulating composition and processing pa- 25M. J. Harrison, A. P. Graebner, W. J. McNeil, and D. S. McGregor, ‘‘Carbon rameters, and have been able to achieve an increase in room Coating of Fused Silica Ampoules,’’ J. Cryst. Growth, 290 [2] 597–601 (2006). temperature resistivity over an order of magnitude greater than 26B. R. Johnson, M. J. Schweiger, and S. K. Sundaram, ‘‘Chalcogenide Nano- previously reported. wires by Evaporation–Condensation,’’ J. Non-Cryst. Solids, 351, 1410–6 (2005). 27A. Hruby and L. Stourac, ‘‘Glass Forming Ability And Electrical Properties Of Semiconducting AsxTe1 x Glasses,’’ Czech. J. Phys., B 24 [10] 1132–8 (1974). 28A. Hruby, ‘‘EvaluationÀ of Glass-Forming Tendency by DTA,’’ Czech. J. Acknowledgments Phys., B22, 1187–93 (1972). 29V. D. Okunev, Y. I. Otman, and A. G. Yurov, ‘‘Properties of Amorphous The authors would like to acknowledge the contributions of Xiangyun Qiu for CdGeAs2 Films Obtained by Cathode Sputtering,’’ Sov. Phys. J., 22 [12] 1328–31 his consultation on how to performing PDF analysis. We would also like to ac- (1979). knowledge Randall Scheele and Anne Kozeliski from PNNL for their assistance in 30V. D. Okunev and N. N. Pafomov, ‘‘Electrical Activity of Ni in Glassy collecting the thermal analysis data. Additionally, the following Department of CdGeAs2,’’ Soviet Physics – Semiconductors, 22 [3] 302–3 (1988). Energy sponsored facilities are acknowledged, as portions of this research were 31I. Kaban, P. Jovari, and W. Hoyer, ‘‘Partial Pair Correlation Functions of performed in them: The Environmental Molecular Synthesis Laboratory, located Amorphous and Liquid Ge15Te85,’’ J. Optoelec. Adv. Mater., 7 [4] 1977–81 (2005). at PNNL, and the Materials Research Laboratory, located at the University of 32A. Pasewicz, D. Idziak, J. Koloczek, P. Kus, R. Wrzalik, T. Fennell, Illinois at Urbana-Champaign. V. Honkima¨ ki, A. Ratuszna, and A. Burian, ‘‘Pari Correlatin Function Analysis of 5-(4-hexadecyloxyphenyl)-10,15,20-tri(4-pyridyl)porphyrin and 5-(4-methoxycar- bonylphenyl)-10,15,20-tri(4-pyridyl)porphyrin,’’ J. Mol. Struct., 875, 167–72 References (2008). 33T. Proffen, S. J. L. Billinge, T. Egami, and D. Louca, ‘‘Structural Analysis of 1J. Mort, The Anatomy of Xerography. McFarland, London, 1989. Complex Materials Using the Atomic Pair Distribution Function—A Practical 2A. Luque and S. Hegedus, Handbook of Photovoltaic Science and Engineering. Guide,’’ Z. Kristallogr., 218, 132–43 (2003). John Wiley and Sons Ltd., Chichester, West Sussex, England, 2003. 34X. Qui, J. W. Thompson, and S. J. L. Billinge, ‘‘PDFgetX2: A GUI Driven 3M. Hoheisel and L. Batz, ‘‘Requirements on Amorphous Semiconductors Program to Obtain the Pair Distribution Function from X-Ray Powder Diffrac- for Medical X-Ray Detectors,’’ J. Non-Cryst. Solids, 266–269 [Part B] 1152–7 tion Data,’’ J. Appl. Cryst., 37, 678–78 (2004). (2000). 35S. F. Marenkin, V. M. Novotortsev, K. K. Palkina, S. G. Mikhailov, and V. T. 4 S. Kasap, J. Rowlands, K. Tanioka, and A. Nathan, ‘‘Applications of Disor- Kalinnikov, ‘‘Preparation and Structure of CdGeAs2 Crystals,’’ Inorg. Mater., 40 dered Semiconductors in Modern Electronics: Selected Examples’’; pp. 149–77 [2] 93 (2004). 36 in Charge Transport in Disordered Solids with Applications in Electronics,Editedby L. Cervinka and A. Hruby, ‘‘The Crystal Structure of CdAs2,’’ Acta Crystal- B. Sergei, John Wiley and Sons Ltd., Hoboken, NJ, 2006. logr. Sect. B, B26, 457–8 (1970). &

8.9

IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 56, NO. 3, JUNE 2009 863 DC Ionization Conductivity of Amorphous Semiconductors for Radiation Detection Applications Bradley R. Johnson, Member, IEEE, Jarrod V. Crum, S. K. Sundaram, Renee M. Van Ginhoven, Member, IEEE, Carolyn E. Seifert, Member, IEEE, Brian J. Riley, and Joseph V. Ryan

Abstract—DC ionization conductivity measurements were used that have not yet been fully resolved. Consequently, a program to characterize the electrical response of amorphous semicon- of research was initiated to examine the feasibility of using ductors to ionizing radiation. Two different glass systems were amorphous semiconductors as the active sensing medium for examined: a chalcopyrite glass ( ; for ) with a tetrahedrally coordinated structure and a chalcogenide radiation detection applications. Two different families of glass ( ; where - ), with a layered or amorphous semiconductors were selected, a chalcopyrite glass three dimensionally networked structure, depending on Te content. family, (CGA) and a chalcogenide glass family, Changes in DC ionization current were measured as a function of . the type of radiation ( or ), dose rate, applied field, specimen Amorphous materials provide several processing advantages thickness and temperature. The greatest DC ionization response was measured with at from an alpha compared to single crystal materials. These include faster syn- source (which is the first reported result for radiation response thesis routes (hours/days vs. weeks to grow an ingot), custom from an amorphous chalcopyrite semiconductor). Avalanche gain shape casting or molding, and the ability to deposit them in was observed in with exposure to alpha radiation at large areas as homogeneous thin films. Additionally, they have fields . These results demonstrate the potential broader compositional flexibility, are less sensitive to trace im- of these materials for radiation detection applications. purities, and their properties can be tailored (within limits) to Index Terms—Amorphous semiconductors, arsenic tri-se- meet specific application needs by controlling processing, com- lenide DC ionization, cadmium germanium di-arsenide, positional, and doping variables. Their main liability, however, . is that they typically have poorer charge carrier transport properties than single crystal semiconductors. This is due to a high I. INTRODUCTION density of localized electronic states near the band edges, which is a natural consequence of their disordered structure. Conduction in amorphous and disordered semiconductors has OR room temperature radiation detection, a semicon- been extensively studied [1]. Much of the early work in the F ductor detector material should have acceptable carrier area of semiconductor theory for amorphous materials was pi- mobilities and lifetimes, moderate band-gaps (e.g., ), a oneered by Mott [2]–[5]. Although considerable emphasis had large absorptive cross-section for interacting with the radiation, been placed on studying doped and hydrogenated amorphous and ideally would be easily synthesizable in large shapes at low silicon [6]–[8], the principles of semiconduction in amorphous cost. Single element, single crystal semiconductors such as Si materials have been applied to chalcogenide and chalcopyrite and Ge have excellent charge carrier mobilities and lifetimes; glasses as well [9], [10]. however, due to their small band gaps, cryogenic cooling is To summarize, the concepts of a conduction and a valence often necessary (depending on the type of detector) to reduce band with an energy (mobility) gap separating them are appli- thermally generated leakage currents below the signal level cable to amorphous semiconductors just as they are for crys- induced by ionizing radiation. The other option to reduce the ef- talline semiconductors. The main distinction is that their in- fects of thermally generated leakage currents is to use materials herent long-range structural disorder results in a localization of with larger band gaps. This necessitates the use of compound the electronic states at the edges of the conduction and valence (binary, ternary, etc.) semiconductors such as CdZnTe (CZT), bands and in some instances within the mobility gap as well. AlSb, HgI, etc. However, growing pure, defect-free single The presence of these localized states necessitates that charge crystal compound semiconductors that meet the geometrical carriers hop from one state to an adjacent state as they move and performance requirements for radiation detector applica- through the material. This tends to impede the transport kinetics tions is a challenging processing problem with many obstacles across the mobility gap and at the edges until the carrier reaches the extended states [10]. The net result is that charge carrier Manuscript received July 12, 2008; revised November 24, 2008. Current ver- transport is impeded. sion published June 10, 2009. This work was supported in part by the Depart- Notwithstanding, these materials have been shown to have ment of Energy, National Nuclear Security Administration, Office of Defense Nuclear Nonproliferation, Office of Nonproliferation Research and Develop- some unique electrical properties that occur at high fields such ment (NA-22). PNNL is operated for the U.S. Department of Energy by Battelle as non-destructive switching, reversible memory and avalanche Memorial Institute under Contract DE-AC06-76RLO 1830. gain as well as a variety of photo-induced phenomena [11]–[13]. The authors are with the Pacific Northwest National Laboratory, 790 6th Street, Richland, WA 99354 USA (e-mail: Bradley. [email protected]). In many chalcogenide glasses (e.g., Se, , , and Digital Object Identifier 10.1109/TNS.2009.2013344 their sulfur analogs), these unique properties can be correlated

Authorized licensed use limited to: Pacific Northwest National Laboratory. Downloaded on8.10 July 10, 2009 at 20:15 from IEEE Xplore. Restrictions apply.

864 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 56, NO. 3, JUNE 2009 to the presence of non-bonded valence pair electrons [14]. Ap- recombination distances raise concerns about charge collection parently, the application of high fields to these materials is able in these compounds. The DC ionization experiment described to create a large population of carriers above the Fermi level below provides an initial view into the charge carrier transport that can subsequently be promoted above their mobility edges mechanisms in amorphous semiconducting materials, and by thermal excitation [15]. enables a straightforward means of comparing the performance Gamma-radiation induced effects in chalcogenides have been of various compounds. studied since the 1980s [16]–[25]. One notable phenomenon is physical ageing [17] at high doses for As-Se chalcogenide II. EXPERIMENTAL PROCEDURES glasses under -irradiation, which was analyzed using a differential scanning calorimetry (DSC) instrument. An in- A. Specimen Synthesis crease in glass transition temperature and the area of an en- Bulk amorphous semiconductor specimens were synthesized dothermic peak near the glass transition region for Se-enriched from high-purity elements ( ; Alpha-Aesar) in the glasses have been associated with additional -activated struc- specified stoichiometric ratios. All elements were stored and tural relaxation of the glass network towards thermodynamic batched in a nitrogen-purged, atmosphere controlled glove box equilibrium as a supercooled liquid [19]. Transient radiation ef- (M.Braun Inc.) where water and oxygen levels were maintained fects have also been reported in amorphous chalcogenide semi- at . Reaction vessels (ampoules) were made from conductors. For instance, gamma-ray induced conductivity had 1-mm thick fused quartz tubing with either 10 or 25 mm in- been demonstrated for some As-Se-Te glasses, and x-ray inside diameters (GE214, GM Associates, Inc.). They were pre- duced photoconductivity was shown for As- Se-Tl glass [26], pared by sealing one end of a long tube with an [27]. oxygen-propane torch. Each tube was cleaned using a RCA1 Because of their unique properties, amorphous semiconduc- etch where the tubes were filled with a solution of tors have been successfully applied in a number of applica- :de-ionized water (DIW) at a 1:1:5 ratio by volume and tions such as direct conversion digital x-ray image detectors set in a 70 oven for 2 hours [36]. The tubes were rinsed with [28], x-ray photoconductors [12], [29], dosimeters [30], photo- DIW and then etched in a 5% HF solution at room temperature voltaics [31], [32], phase change memory storage devices and for 2 hours [37]; the tubes were rinsed again with DIW and then photocopiers [33], [34]. However, they have not been widely were baked out and annealed at 1160 inside of a box furnace developed for radiation detection applications. This study is an located within the atmosphere controlled glovebox. initial investigation to evaluate radiation response properties of Stoichiometric quantities of the elemental constituents were selected amorphous semiconductors from two different glass weighed out to accuracy and loaded into the ampoules. structure systems. Each batch contained of raw elements so as to Radiation detection depends not only on materials perfor- yield an ingot with a 10 mm diameter that was 30–40 mm tall. mance, but also on the electrical engineering and signal pro- Detailed information regarding synthesis and process steps to cessing involved with extracting and processing the electrical batch and seal the ampoules are provided elsewhere [38], [39]. response of the material. Since each material will have different Once constructed, the ampoules were attached to a holder charge carrier dynamics and transport properties, it is not nec- and loaded into the containment assembly of a custom de- essarily possible to simply insert a new material into an existing signed rocking furnace. Specimens were thermally processed detector assembly and evaluate its performance. Thus, ideally, by slowly heating them up in the rocking furnace up to 850 there would be a technique that could be used to efficiently and rocked for 24 h. After thermal processing, the rocking screen the performance of new materials. In light of this need, motion was turned off to allow the liquid to settle to the bottom Derenzo and coworkers [35] developed an apparatus for con- of the ampoule. The furnace temperature was then reduced fining powder samples under pressure in an electric field and to in preparation for quenching (varied with composition). measuring the DC current induced by a 450 Ci, 1.2 MeV Then the ampoules were quickly extracted from the furnace gamma source at a dose rate of 1,500 rad/min. This approach and quenched in water so as to vitrify the liquid and prevent has been adapted to screen and evaluate bulk amorphous semi- crystallization. After quenching, the ingots were annealed at conductors for their potential application as radiation detection less than for that composition (as determined materials. by thermal analysis experiments and extrapolation from litera- We have extended the argument originally developed by ture data) for hours to aid in reducing thermally-induced Derenzo and coworkers [35] for crystalline semiconductor stress in the glasses. After cooling, the ingots were cast into powders to bulk amorphous semiconductors. Similar to poly- an acrylic resin (Buehler VariKleer®), sectioned into 1–3 mm crystalline semiconductor powders, trapping, detrapping and thick wafers, and polished. retrapping are expected to occur in amorphous semiconductors as the moving charge carriers interact with structural disorders, B. DC Ionization Measurements impurities, and defects. This will add to the effective carrier After polishing the specimens, electrodes were applied using lifetime, . Nonetheless, many of the carriers will ultimately a shadow mask and a vapor deposition process. A variety of recombine in the material, and will not traverse the complete different electrodes such as evaporated carbon, sputtered Pd, distance between the electrodes. Thus, the average drift distance and sputtered Au/Pd were evaluated. Some experiments were between trapping and recombination will be much less done without electrodes, per se, and direct electrical contact than the distance between electrodes (d). Potentially shorter was made to the specimen using either conductive rubber or

Authorized licensed use limited to: Pacific Northwest National Laboratory. Downloaded on8.11 July 10, 2009 at 20:15 from IEEE Xplore. Restrictions apply.

JOHNSON et al.:DCIONIZATIONCONDUCTIVITYOFAMORPHOUSSEMICONDUCTORSFORRADIATIONDETECTIONAPPLICATIONS 865

Fig. 2. DC ionization current response for from exposure to a 3.9 mCi -source. Fig. 1. Current-voltage curve for glass at , with and without exposure to a 3.9 mCi source.

III. RESULTS conductive, adhesive carbon tape. The different types of elec- A. trodes were evaluated by measuring their current-voltage (I–V) response curves. All of the electrodes analyzed showed an Current-voltage curves up to a bias voltage of 100 V were col- ohmic response, with only nominal differences in performance lected from a 2.3 mm thick specimen of CGA glass with surface between them. The specimens were then placed in a sample contacts at with and without exposure to the source holder, and electrical contact was made to the electrodes using (Fig. 1). It was necessary to cool the detector due to high leakage either conductive tape, silver epoxy or spring loaded metal pins. currents at room temperature. Direct electrical contact (without Bias voltages were applied to the specimens using source electrodes) was made to the specimen using silver epoxy and measurement unit (SMU, e.g., either a Keithley 237, 6430, or conductive, adhesive carbon tape. Both contacts were applied to 617, depending on the experiment). These devices were capable the front face of the specimen. A digital data acquisition system of simultaneously sourcing voltage, and measuring current. For was used to control the voltage ramp rate and record the I–V a given bias voltage, the current flowing through the device data at each point. The same ramp rate profile was used for both was measured with and without exposure to a sealed radiation measurements. A net DC ionization current of 36.4 nA was mea- source. The difference between current readings was called sured at 100 V. This represented approximately a 50% increase the DC ionization current, . Alpha radiation response in conduction compared to the dark current. measurements were made using a 3.9 mCi source and radiation response measurements were made with a 20 Ci B. sealed source. The dose rate from the source was Additional DC ionization experiments were done with a 3.44 varied by controlling the source-specimen separation distance. mm thick vitreous specimen of As-Se-Te (Fig. 2). Direct elec- Due to equipment availability limitations, and specimen-spe- trical contact was made to the specimen using conductive, adhe- cific testing requirements, a variety of different experimental sive carbon tape. Experiments were done with both electrodes configurations were used. on the front face of the specimen, as well as with electrodes on Detector gain was calculated according to (1) [33], [34]. opposing faces. Measurements were made at room temperature at two different bias voltages: 500 and 750 V. A movable shutter was placed between the specimen and the sealed source and the specimen was biased up to the specified voltage. The IV data (1) was logged using a computer controlled data acquisition system. While under bias, the shutter was manually moved to expose the where is the percentage of detector gain, is the bias specimen to the source, and the corresponding change in current voltage, is the DC ionization current, is the activity of was recorded. At 500 V, the DC ionization current was 21.5 nA, the source in decays per second, and is the average energy while at 750 V, the ionization current was 27.8 nA. This repre- of the particle at the face of the detector as compensated for sented a current increase of 41.8% and 34.9%, respectively. attenuation through air and the foil seal. Note that this is a simple ratio of the electrical power readout from the source C. measuring unit as compared to the power (J/s) deposited onto Results from DC ionization experiments on chalco- the face of the detector element by the alpha source. genide glass conducted with a sealed source are shown in

Authorized licensed use limited to: Pacific Northwest National Laboratory. Downloaded on8.12 July 10, 2009 at 20:15 from IEEE Xplore. Restrictions apply.

866 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 56, NO. 3, JUNE 2009

Fig. 3. Variation in DC ionization current with applied bias voltage for Fig. 5. Comparison of DC ionization response for various amorphous semicon- from exposure to a 3.9 mCi source. ductors from exposure to a 3.9 mCi -source. ( front face contacts; front face contacts; opposite face contacts; opposite face contacts).

energy/decay) by the source. The increase in detector gain as a function of applied field was an indication that the specimen was undergoing avalanche gain. The relative performance of the three different specimens ( , , and ) under exposure to the sealed alpha source is shown in Fig. 5. As determined by the different experimental apparatuses used, some measurements were done with both contacts applied to the front face of the specimen (1, 2), and for other measurements, the contacts were applied to opposing faces (3, 4). DC ionization experiments were also conducted using a 20 Ci -source on . The DC ionization current was measured through the bulk of the specimen as a function of the applied field, specimen thickness, and the dose rate. The data Fig. 4. Detector gain as a function of applied field for from exposure is plotted in Fig. 6. The DC ionization current was observed to 3.9 mCi -source. to increase with each of these variables: field strength, sample thickness, and dose rate. Figs. 3 and 4. The specimen was 838 thick. Measurements were made through the bulk of the specimen (contacts on oppo- IV. DISCUSSION site faces of the specimen) both with and without evaporated Three different amorphous semiconductors were evaluated planar carbon electrodes (negligible difference was observed from two different glass families, and each of them showed a with and without electrodes). Electrical contact was made using measurable response to ionizing radiation from an -source. conductive, adhesive carbon tape. Measurements were made at The specimen showed a response to both and ra- room temperature at a variety of different bias voltages using the diation (the only one evaluated with a -source so far). The same procedure as for the As-Se-Te specimen. The variation of CGA glass specimen from the chalcopyrite family (which re- the DC ionization current as a function of the applied field is quired modest cooling) was potentially the most sensitive mate- shown in Fig. 3. rial evaluated, showing a 50% increase in DC ionization current These data were used as per (1) to calculate the percentage of at only 100 V of bias (Fig. 1) and having the largest DC ioniza- detector gain as a function of the applied field. The electric field tion response as a function of applied field (Fig. 5). was assumed to be uniform across the thickness of the sample. All of the specimens showed a prompt (e.g., ) response The results are shown in Fig. 4. There was a slight asymmetry in upon changing their exposure to the sealed source. This indi- the data, with the negatively applied biases yielding a margin- cates that when the specimen was exposed to the source, the ally greater response. At a field of , the de- traps in the sample started to fill as the current increased sharply tector gain was approximately 280%. This means that the power to a new equilibrium level. At this point, the rate of generation output of the detector at this field was a factor of 2.8 of electron-hole pairs was equal to the rate of trapping and re- greater than the power deposited into the specimen (activity combination. When the shutter was placed between the source

Authorized licensed use limited to: Pacific Northwest National Laboratory. Downloaded8.13 on July 10, 2009 at 20:15 from IEEE Xplore. Restrictions apply.

JOHNSON et al.:DCIONIZATIONCONDUCTIVITYOFAMORPHOUSSEMICONDUCTORSFORRADIATIONDETECTIONAPPLICATIONS 867

significant role that chemistry plays in the performance of amorphous semiconductors, and indicates that potentially greater im- provements are possible through optimized materials design. Gamma radiation experiments on (Fig. 6) also showed that the DC ionization response varied with dose rate, which was anticipated with an increased flux of ionizing photons. The other important observation was that the DC ionization response also increased with specimen thickness. The same bias voltage was applied to two specimens with different thicknesses, thereby creating two different field densities. The thicker specimen with the lower electric field had the larger DC ionization response, thus illustrating the fact that detector efficiency increases with volume.

V. C ONCLUSIONS The experimental techniques explained in this paper for measuring the DC ionization currents proved to be a useful Fig. 6. Variation in DC ionization current with applied field, dose rate, and characterization tool for evaluating the radiation response of specimen thickness for from exposure to a 20 Ci source. amorphous semiconductors. Both of the amorphous semiconductor families evaluated (chalcopyrite and chalcogenide) showed a measurable electrical response to ionizing radiation, and the specimen, the current dropped sharply to the initial level thus demonstrating their potential use for radiation detector as the charge carriers detrapped. Multiple cycles showed prac- applications. The material with the strongest response to -ra- tically no drift in the baseline current level (Fig. 2). diation was the glass specimen at .To Increasing the applied field resulted in an increase in the DC the best of our knowledge, these are the first reported results of ionization current response. This was observed with exposure -radiation induced conductivity in an amorphous chalcopyrite to both and radiation (Fig. 3 through Fig. 6). The effect was semiconductor. The specimen had a slightly such that trapped and lost charge carriers could not only be com- lower DC ionization response than the CGA glass, but it could pensated for, but that additional charge carriers could also be be operated at room temperature, and the response increased generated. There are a couple of possible explanations for the with the applied field. Particularly noteworthy is the observa- observed behavior. With increasing field strength, the moving tion that the incorporation of Te in As-Se glass produced a charge carriers might have sufficient momentum to escape low four-fold increase in DC ionization response as compared to energy trap states that might otherwise capture them. It is also pure arsenic triselenide glass. The specimen could possible that with increasing field strength, once trapped charge also be operated at room temperature, and showed a DC ion- carriers could be released from low energy trap states and par- ization response to both – and -radiation. Additionally, due ticipate in conduction process again. A third option is that at to its high resistivity, it could be operated at fields in excess extreme fields, the moving charge carriers could be accelerated of without breaking down. At fields in excess of with sufficient momentum so as to be able to create additional , avalanche gain was observed, such that the electron-hole pairs through impact ionization as they encounter power generated from the device exceeded the power deposited defects, trap sites, or stationary charge carriers. This appears to into it by the radiation source. Detector gain values as high as be the dominant mechanism at extreme fields as shown in the 280% were measured at field of . specimen exposed to the alpha source (Fig. 4). Im- pact ionization was able to create an avalanche of additional ACKNOWLEDGMENT charge carriers that compensated for trapping losses at a field The authors would like to gratefully acknowledge the assis- of and subsequently created a detector gain of tance of C. Chamberlain and Y. Zhang for their assistance on 280% at a field of . Impact ionization induced this project. avalanche has been reported in other amorphous semiconductors such as amorphous selenium [12], so this appears to be a REFERENCES reasonable explanation for the observed behavior. The transi- [1] B. Pistoulet, J. L. Robert, J. M. Dusseau, and L. Ensuque, “Conduction tion between different charge carrier transport mechanisms may mechanisms in amorphous and disordered semiconductors explained by a model of medium-range disorder of composition,” J. Non-Cryst. be difficult to determine experimentally, since the detector gain Solids, vol. 29, pp. 29–40, 1978. with applied field is a continuous, smooth curve. This is prob- [2] N. F. Mott, “Introductory talk: Conduction in non-crystalline mate- ably due to the fact that the density of defect states as a function rials,” J. Non-Cryst. Solids, vol. 8–10, pp. 1–18, 1972. [3] N. F. Mott, “The origin of some ideas on non-crystalline materials,” J. of energy in these materials also tends to be fairly continuous. Non-Cryst. Solids, vol. 28, pp. 147–158, 1978. The addition of Te to As-Se glass was shown to have a sig- [4] N. F. Mott, “Closing address,” J. Non-Cryst. Solids, vol. 35–36, pp. nificant impact on its charge carrier transport properties. The 1321–1323, 1980. [5] N. F. Mott, “Pre-exponental factor in the conductivity of liquid and DC ionization response for the glass was almost amorphous semiconductors,” J. Non-Cryst. Solids, vol. 77–78, pp. four times greater than that for . This demonstrates the 115–120, 1985.

Authorized licensed use limited to: Pacific Northwest National Laboratory. Downloaded8.14 on July 10, 2009 at 20:15 from IEEE Xplore. Restrictions apply.

868 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 56, NO. 3, JUNE 2009

[6] N. F. Mott, “Conduction bands in a non-crystalline environment,” J. [24] O. I. Shpotyuk, V. O. Balitska, and M. M. Vakiv, “Effect of elec- Non-Cryst. Solids, vol. 90, pp. 1–8, 1987. tron-induced dichroism in vitreous ,” J. Non-Cryst. Solids, vol. [7] N. F. Mott, “Physical aspects of localization,” J. Non-Cryst. Solids, vol. 227–230, pp. 837–841, 1998. 97–98, pp. 213–215, 1987. [25] O. I. Shpotyuk and A. O. Matkovskii, “Radiation-stimulated processes [8] N. F. Mott, “Conduction at a mobility edge,” J. Non-Cryst. Solids, vol. in vitreous arsenic trisulphide,” J. Non-Cryst. Solids, vol. 176, pp. 97–98, pp. 531–538, 1987. 45–50, 1994. [9] H. Fritzsche, “Electronic phenomena in amorphous semiconductors,” [26] B. T. Kolomiets and V. M. Lyubin, “Photoelectric phenomena in amor- Annual Review of Materials Science, vol. 2, pp. 697–744, 1972. phous chalcogenide semiconductors,” Phys. Stat. Sol. (a), vol. 17, pp. [10] P. Nagels, “Electronic transport in amorphous semiconductors,” in 11–46, 1973. Amorphous Semiconductors, M. H. Brodsky, Ed., 2nd ed. New York: [27] T. Minami, A. Yoshida, and M. Tanaka, “Gamma-ray induced conduc- Springer-Verlag, 1979, vol. 36, pp. 113–158. tivity of vitreous semiconductors in the systems As-S-Te and As-Se- [11] S. R. Ovshinsky and H. Fritzsche, “Amorphous semiconductors for Te,” J. Non-Cryst. Solids, vol. 7, pp. 328–36, 1972. switching, memory and imaging applications,” IEEE Trans. Elect. [28] M. Hoheisel and L. Batz, “Requirements on amorphous semiconduc- Dev., vol. ED-20, pp. 91–105, 1973. tors for medical X-ray detectors,” J. Non-Cryst. Solids, vol. 266–269, [12] S. Kasap, J. Rowlands, K. Tanioka, and A. Nathan, “Applications of pp. 1152–1157, 2000. disordered semiconductors in modern electronics: Selected examples,” [29] S. Ross, G. Zentai, K. S. Shah, R. W. Alkire, I. Naday, and E. M. West- in Charge Transport in Disordered Solids with Applications in Elec- brook, “Amorphous silicon, semiconductor X-ray converter detectors tronics, B. Sergei, Ed., 2006, pp. 149–177. for protein crystallography,” Nuclear Instrum. Meth. Phys. Res. A (Ac- [13] A. V. Kolobov and K. Tanaka, “Photoinduced phenomena in amor- celerators, Spectrometers, Detectors and Associated Equipment), vol. phous chalcogenides: From phenomenology to nanoscale,” in Hand- 399, pp. 38–50, 1997. book of Advanced Electronic and Photonic Materials and Devices, H. [30] G. A. M. Amin and N. M. Spyrou, “Study of gamma-radiation-induced S. Nalwa, Ed. San Diego, CA: Academic, 2001, vol. 5, Chalcogenide optical effects in Ge-Se-Cd for possible industrial dosimetric applica- Glasses and Sol-Gel Materials, pp. 47–90. tions,” Rad. Phys. Chem., vol. 72, p. 419, Mar. 2005. [14] A. Kolobov, H. Oyanagi, A. Roy, and K. Tanaka, “Role of lone-pair [31] L. C. Burton, C. F. Smith, Jr, A. F. Carroll, and L. H. Slack, “Studies electrons in reversible photostructural changes in amorphous chalco- of for photovoltaic applications,” in Proc. Southeastcon’79, genides,” J. Non-Cryst. Solids, vol. 227–230, pp. 710–714, 1998. Roanoke, VA, 1979, pp. 25–28. [15] H. Fritzsche and S. R. Ovshinsky, “Electronic conduction in amor- [32] S. Kumar, B. R. Mehta, S. C. Kashyap, and K. L. Chopra, “Amorphous phous semiconductors and the physics of the switching phenomena,” chalcogenide thin-ﬁlm schottky barrier solar cell,” J. Non-Cryst. Solids, vol. 2, pp. 393–405, 1970. Appl. Phys. Lett., vol. 52, pp. 24–26, 1988. [16] V. Balitska, O. Shpotyuk, and H. Altenburg, “Bimolecular relaxation [33] J. Schottmiller, M. Tabak, G. Lucovsky, and A. Ward, “The effects of kinetics observed in radiation-optical properties of Ge-As(Sb)-S valency on transport properties in vitreous binary alloys of selenium,” glasses,” J. Non-Cryst. Solids, vol. 352, pp. 4809–4813, 2006. J. Non-Cryst. Solids, vol. 4, pp. 80–96, 1970. [17] R. Y. Golovchak, A. Kozdras, C. Gorecki, and O. I. Shpotyuk, [34] M. D. Tabak, “High ﬁeld transport and impurity effects in amorphous “Gamma-irradiation-induced physical ageing in As-Se glasses,” J. Se and As-Se alloy system,” in Proc. 2nd Int. Conf. Conduction Low Non-Cryst. Solids, vol. 352, pp. 4960–4963, 2006. Mobility Materials, Eliat, Israel, 1971, p. 276. [18] O. Shpotyuk, A. Kozdras, T. Kavetskyy, and J. Filipecki, “On the [35] S. Dorenzo, E. Bourret-Courchesne, F. J. James, M. K. Klintenberg, Y. correlation between void-species structure of vitreous arsenic selenide Porter-Chapman, J. Wang, and M. J. Weber, “Identifying semiconduc- studied with X-ray diffraction and positron annihilation techniques,” tors by D.C. ionization conductivity,” in 2005 IEEE Nuclear Science J. Non-Cryst. Solids, vol. 352, pp. 700–703, 2006. Symposium Conference Record, Puerto Rico, 2005, pp. 1132–1134. [19] O. I. Shpotyuk and R. Y. Golovchak, “Physical aging effects in selenide [36] Kern and Duotinen, RCA Review, vol. 31, p. 186, 1970. glasses accelerated by highly energetic [gamma]-irradiation,” J. Non- [37] B. J. Riley, B. R. Johnson, S. K. Sundaram, M. H. Engelhard, R. E. Cryst. Solids, vol. 352, pp. 704–708, 2006. Williford, and J. D. Olmstead, “Pressure-temperature dependence of [20] O. I. Shpotyuk, A. P. Kovalskiy, T. S. Kavetskyy, and R. Y. Golovchak, nanowire formation in the arsenic-sulfur system,” Phys. Chem. Glass, “Threshold restoration effects in [gamma]-irradiated chalcogenide vol. 47, pp. 675–80, 2006. glasses,” J. Non-Cryst. Solids, vol. 351, pp. 993–997, 2005. [38] B. R. Johnson, M. J. Schweiger, and S. K. Sundaram, “Chalcogenide [21] V. Balitska, R. Golovchak, A. Kovalskiy, E. Skordeva, and O. I. Sh- nanowires by evaporation-condensation,” J. Non-Cryst. Solids, vol. potyuk, “Effect of Co60 [gamma]-irradiation on the optical properties 351, pp. 1410–1416, 2005. of As-Ge-S glasses,” J. Non-Cryst. Solids, vol. 326–327, pp. 130–134, [39] B. R. Johnson, B. J. Riley, S. K. Sundaram, J. V. Crum, C. Henager, 2003. Y. Zhang, V. Shuthanandan, C. E. Seifert, R. M. V. Ginhoven, and [22] O. I. Shpotyuk, J. Filipecki, A. Kozdras, and T. S. Kavetskyy, “Radia- C. Chamberlin, “Synthesis and characterization of bulk vitreous tion-induced defect formation in chalcogenide glasses,” J. Non-Cryst. ,” J. Amer. Cer. Soc., 2009. Solids, vol. 326–327, pp. 268–272, 2003. [23] V. O. Balitska and O. I. Shpotyuk, “Radiation-induced structural trans- formations in vitreous chalcogenide semiconductors,” J. Non-Cryst. Solids, vol. 227–230, pp. 723–727, 1998.

8.15 Authorized licensed use limited to: Pacific Northwest National Laboratory. Downloaded on July 10, 2009 at 20:15 from IEEE Xplore. Restrictions apply.

J. Am. Ceram. Soc., 95 [7] 2161–2168 (2012) DOI: 10.1111/j.1551-2916.2012.05171.x Published 2012. This article is a U.S. Government work and is in the public domain in the USA. Journal Tricadmium Digermanium Tetraarsenide: A New Crystalline Phase Made with a Double-Containment Ampoule Method

Brian J. Riley,†,* Bradley R. Johnson,* Jarrod V. Crum,* and Michael R. Thompson Paciﬁc Northwest National Laboratory, Richland, WA 99352, USA

A new crystalline phase in the Cd-Ge-As family of materials II. Background has been recently discovered. This phase was made with a One of the more widely studied Cd Ge As compounds is unique double-containment ampoule method that provided a x y z the stoichiometric AIIBIVC V chalcopyrite (hereafter referred specific cooling rate along with a thermal compression mecha- 2 to as 1-1-2). Historically, the 1-1-2 compound is most often nism. The composition of this crystalline phase was determined preferred in single-crystal form, however, the growth of large to be Cd Ge As with energy dispersive spectroscopy. The 3 2 4 single crystals has proven challenging.12 The consensus has detailed methods used to fabricate these crystals are presented. been that the anisotropic thermo-physical properties of this The crystal structure is still under investigation, although preli- material are the root cause for these problems. The large minary results are given. variations in the linear coefficients of thermal expansion between the a-axis (aL(a)) and the c-axis (aL(c)) tend to create I. Introduction strain gradients that lead to defects and cracking in large bo- ules.9,17,27–29 Most of the early work with this material ERNARY chalcopyrites are a family of materials with com- revealed issues with polycrystallinity, cracking, and poor position I-III-VI (AIBIIIC VI)orII-IV-V (AIIBIVC V) T 2 2 2 2 optical properties, although recently, new processing tech- that can be formed as amorphous, polycrystalline, and/or niques were developed to produce high purity single crystals single-crystal compounds that are often semiconductors. The using a horizontal gradient freeze growth method.11,22,26,30,31 II-IV-V chalcopyrites are analogous to the widely studied 2 To determine the best processing route to create large sin- III-V materials (i.e., GaP, GaAs), except that their crystal gle crystals, detailed phase diagrams and glass-formation structure is tetragonally distorted from the cubic III-V zinc- studies were conducted by implementing melt-quench tech- blende structure.1,2 Numerous applications and important niques where the processing temperatures (T ), and quen- properties have been discovered for chalcopyrite com- p chant temperatures (T ) were similar to those in this current pounds and nonchalcopyrite Cd Ge As compounds includ- q x y z work.1,20–26 One of the more complete phase diagrams was ing photovoltaics and photoconductivity,3–5 semiconductor published by Borshchevskii and Roenkov.22 They studied radiation detection,5,6 spintronics,7 nonlinear optics,8–10 super- CdGe As compounds with a melt-quench method and deter- conductivity,11,12 high refractive index (n)(n 3.5),8,13 high y 2 mined that the 1-1-2 compound coexisted with Ge, GeAs, room temperature electron mobility,14,15 and! windows for CdAs , and Cd As . With a melt-quench approach, Pamplin infrared lasers.8,16,17 2 3 2 and Feigelson determined that at 550°C, the 1-1-2 compound Here, we report the discovery of a new, previously unre- was in equilibrium with Ge, GeAs, CdAs , Cd GeAs , and ported Cd-Ge-As phase with the simplified composition of 2 2 4 Cd As .25 The coexistence of these additional species with Cd Ge As (hereafter referred to as 3-2-4) with a technique 3 2 3 2 4 similar thermodynamic stability increases the difficulty of termed the double-containment (DC) ampoule method. In growing single crystals of the 1-1-2 compound. However, 2008,18 this 3-2-4 compound was first discovered while despite all of this prior work with Cd-Ge-As compounds, attempting to manufacture a variety of bulk, amorphous these prior studies did not report the formation of the 3-2-4 CdGe As compounds (i.e., y = 0.45, 0.65, 0.85, and 1.00).19 y 2 phase as discussed here according to either their reported The 3-2-4 crystals have yet to be created as large, single crys- X-ray diffraction (XRD) patterns or their morphological tals, but they have been repeatedly synthesized under differ- descriptions. One of these studies even batched the ent conditions. Cd Ge As composition directly, but did not report the Here, we present the chemical analysis of the 3-2-4 phase 3 2 4 phase we are presenting here.25 as well as the results from a series of experiments performed Figure 1 summarizes the Cd-Ge-As phases present in the to better understand the formation of this phase. The Cd- International Crystal Structure Database along with the dif- Ge-As ternary diagram has been researched in the past, how- fraction pattern of the 3-2-4 phase and Fig. 2 shows where ever, this particular phase has never been observed. Thus, we the 3-2-4 phase fits into the abridged Cd-Ge-As ternary dia- present a detailed description of the techniques used to fabri- gram.22–25 Although the composition of the 3-2-4 phase is cate these crystals so that these methods are well-docu- close to that of the Cd Ge As (4-3-5) phase reported by mented.1,20–26 In addition, the DC processing method might 4 3 5 Pamplin and Feigelson,25 the diffraction pattern for the 3-2-4 be useful to create other AII-BIV-CV materials of interest to phase did not match that of the 4-3-5 phase, or any other the scientific community, for example, Cd-Sn-As. phase in the International Crystal Structure Database (Fig. 1). Thus, the 3-2-4 materials were deemed unique. A summary of the various Cd-Ge-As compounds that we found in the literature are presented in Table I.23–25,32,33 Hong et al.23 performed melt-quench experiments with J. Heo—contributing editor CdGeyAs2 compounds (where y = 0.3 and 0.6) and reported crystalline phases observed with XRD that they were unable to match to known compounds in the crystal structure data-

Manuscript No. 30484. Received October 12, 2011; approved February 20, 2012. base. The temperature range at which Hong et al. observed *Member, The American Ceramic Society these phases was T ~ 407°C–413°C and the diﬀraction †Author to whom correspondence should be addressed. e-mail: [email protected] 2161

8.16

2162 Journal of the American Ceramic Society—Riley et al. Vol. 95, No. 7

Table I. Summary of Crystalline Cd-Ge-As (CGA) Phases Observed in the Literature (Data are Sorted by Cd at.% from Lowest to Highest)

Simplified Elemental %’s 3-2-4 (Cd-Ge-As) (Cd-Ge-As) Reference(s) (h k l phase) 3.5-45.5-1 7-91-2 Hong et al.23 4.3-27-2 13-81-6 Hong et al.23 1-1-2 1-1-2 25-25-50 Pfister32 4-1-9.3 28-7-65 Hong et al.23 2-1-4 29-14-57 Mikkelson and Hong24, Chernov et al.33 2-1-4 3-1-5.1 33-11-56 Hong et al.23 4-3-5 33-25-42 Pamplin and Feigelson25 3-2-4 33-22-45 Current work Intensity, A.U. Intensity, 4-3-5 II IV V In addition to crystalline ternary A B C2 chalcopyrites, many have worked on exploring the amorphous analogs. In contrast to most oxide glasses, the amorphous structure of Hong chalcogenide- and chalcopyrite-compounds is thought to closely resemble that of the crystalline analogs.23 The most stable ternary amorphous compounds were found to be CdGeAs2, CdSiAs2, ZnGeAs2, and ZnSiAs2 listed in order of decreasing glass-forming tendency (Kgl) based on the Hruby 2T, Degrees criterion.34 Sharma et al.1 explains the Hruby criterion as Fig. 1. Summary of diffraction patterns from the literature Kgl Tc Tg = Tm Tc (1) compared with the calculated pattern (hkl phase, see text) for the ¼ð À Þ ð À Þ 3-2-4 phase. Literature data consists of the 1-1-2,42 2-1-4,24,33 and 4-3-525 phases along with an extracted data set of unknown where T , T , and T are the glass transition, crystallization, 23 g c m diffraction peaks from Hong et al. and solidus or melting temperatures, respectively. A higher Kgl means that a glass is easier to quench into the amorphous state than a glass with a lower Kgl. Glasses with low Kgl values (<1) are poor glass-formers and require fast quenching rates, rc, in order to be supercooled into an amorphous state.35,36 Thus, Boltovets et al. investigated several different quenchants including: H2O and NaCl/H2O(rc ~ 140 –150°C/s at 35°C), liquid gallium (rc ~ 80°C/s at 210°C and 35 rc ~ 55°C/s at 290°C), and liquid tin (rc ~ 40°C/s at 370°C). For the systems with poor glass formation, where Kgl < 0.2, the quench rate required for glass formation was rather high, that is, >50–200°C/s. This quenching rate falls between the amorphous chalcogenides that require low quench rates and amorphous metals that require extremely high quench rates, typically made only in thin film form.35,36 Some of these different quenchants were used in the work presented here.

III. Methods (1) Materials Processing The six samples discussed here were labeled with CGA-# where the # does not necessarily denote the sequential order in which the experiments were conducted. The full process by which the DC ampoules were prepared can be found in our previous work, but a condensed version will also be presented here.19 Two diﬀerent sizes of GE214 fused quartz (GM Associates, Inc., Oakland, CA) reaction vessels were used for these experiments, 10 mm 9 12 mm tubes and Fig. 2. Cd-Ge-As ternary diagram showing the placement of the 19 mm 9 22 mm tubes (inner 9 outer diameter). Each vessel 3-2-4 phase (●). This diagram was recreated from the diagram was prepared similarly where it was: (1) shaped with a torch, presented by Borshchevskii and Roenkov22 with added phases from 25 24 (2) cleaned with an RCA1 etch process at 70°C (2 h) and Pamplin and Feigelson as well as Mikkelsen and Hong (○). The 37 phases from Hong et al. are Ge-deﬁcient (Ge 11 at.%) and are rinsed with deionized water (DIW), (3) soaked in a solution 23 not included here. The top portion shows where the inset (bottom of 5:5:90 HNO3:HF:DIW (by volume) for 2 h and rinsed in 29 portion) is located on the full ternary diagram. DIW, (4) annealed at 1160°C for 1 h in a horizontal quartz tube furnace, and then (5) baked at 200°C in a furnace (Del- patterns are similar to those seen for the 3-2-4 phase tech, Inc. Denver, CO) within a nitrogen glove box, observed in the current study (Fig. 1). However, Hong et al. with < 0.1 ppm of O2/H2O (M-Braun, Inc., Stratham, NH). showed these crystallites (of unknown composition) as being Next, ultra-high purity elemental constituents of Cd (6N), very small (~400 nm 9 1 lm) and irregularly shaped, which Ge (6N), and As (>7N) (Alfa Aesar, Ward Hill, MA) were does not coincide with the oriented morphology of the 3-2-4 added into the inner vessel (10 9 12 mm) while inside the crystallites observed in the current study. glove box. The total mass of the charge, m , varied by experi- s 8.17

July 2012 Cd3 Ge2 As4: A New Crystalline Cd-Ge-As Phase 2163

Table II. Processing Data and Observations for Each Experiment

Sample ID Comp. Method Pi (Pa) Tp (°C) Tq (°C) Tan (°C) mS (g) mCu (g)

4 CGA-1 1-1-2 SC-C 1.2 9 10À 870 W, 20 ± 2 N/A 10.8572 N/A 5 CGA-2 1-1-2 SC 3.4 9 10À 800 Ga, 150 340 13.3010 N/A 5 CGA-3 1-1-2 DC 1.0 9 10À 800 W, 20 ± 2 395 13.3143 24.0 CGA-4 1-1-2 DC 1.5 9 104 825 W, 20 ± 2 360 7.3684 12.5 CGA-5 1-1-2 DC 1.6 9 104 850 W, -2 ± 0.2 360 13.3853 17.8 CGA-6 3-2-4 DC 1.6 9 104 850 W, 20 ± 2 360 13.3426 17.7

All were heat-treated at the processing temperature, Tp, for 25 h and annealed at the annealing temperature, Tq, for 8 h. Pi dictates the initial pressure (in Pa) inside the ampoule containing the elemental constituents (Figure 3(a)). Tp denotes the processing temperature of the ampoule (in °C). Quenchants used were liquid gallium, or Ga, at 150 ± 5°C, and water, or W, at the temperature listed, i.e., 2°C or 20 ± 3°C. The sample mass, m , and mass of copper used, m , are listed À S Cu (in g).

ment with a range of ~7–13 g (Table II). The vessel loaded thermal expansion for copper (aL,Cu) over fused quartz 39–41 with chemicals was transferred to a vacuum assembly where (aL,FQ). 5 7 it was evacuated to 10À Pa (~10À Torr) with a turbomolec- The outer ampoule was evacuated in the same fashion as ular pump backed by a scroll pump. The vessel was purged the inner ampoule and then sealed with a torch. Then, the with semiconductor-grade (ULSI 6N purity) 2.6% H2/Ar- sealed ampoule (Fig. 3(b)) was instrumented with type-K balance (Matheson Tri-Gas, Newark, CA) and then sealed thermocouples (OMEGA Engineering, Inc., Stamford, CT) under vacuum with a torch. This concludes the preparation that were connected to a data acquisition system (HYDRA for the single-containment ampoule (SC) as seen in Fig. 3(a). 2620, Fluke Corporation, Everett, WA) for temperature For one of the specimens discussed here, CGA-1, the inte- monitoring during the heat-treatment process. The ampoule rior of the SC vessel was coated with a pyrolytic graphite with thermocouples was connected to a stainless steel basket layer by covering the interior of the ampoule with acetone and inserted into an Inconel® secondary containment vessel (99.9% purity, Fisher Scientific, Pittsburgh, PA), pouring it spanning through the hot-zone region of a Deltech furnace out, and then heating the vessel with a torch while it was on an adjustable, custom-built rocking platform. The furnace inverted. This was done as an extra step prior to inserting was set to rock with an 18.5-cm radius and pause at the the charge to help prevent chemical attack on the quartz by extremes of a 30° throw (from horizontal) for 15 s dwell the charge; however it was later deemed unnecessary as the times at each extreme with a heating profile as follows: chemicals did not noticeably attack the quartz.19 The method of implementing carbon-coated SC ampoules was termed the 3C=min 2h 3C=min 20C 400C 400C 625C “SC-C” method (Table II). ! ! ! For the four DC experiments presented in Table II, once 2h 3C=min 25 h the ampoule with the elements was vacuum-sealed constitut- 625C Tp Tp quench Tq ! ! ! ! ð Þ ing the SC ampoule, this ampoule was then centered verti- (2) cally inside of the 19 9 22 mm tube and -75 lm (-200 mesh) copper powder was loaded into the annulus between the ves- T ° sels. The height of copper powder varied by experiment, but Samples were held around the m of cadmium (321 C) and ° was poured to match the predicted height of the final melted sublimation temperature of arsenic (616 C) to allow for slow charge. This height was calculated with the measured density mixing to control the vapor pressures inside the ampoule(s). of amorphous CdGeAs , ~5.74 g/cm3 presented in the litera- Following the heat-treatment, the sample holder and 2 ampoule were quickly removed from the furnace and ture, and the total mass of copper, mCu, used in each experi- 19,38 quenched immediately into either water (T = 2 ± 0.2°C or ment is listed in Table II. The copper charge was added q À as a thermal conductor between the two quartz tubes and is 20 ± 2°C) or liquid gallium (Tq = 150 ± 5°C), where the explained in further detail in our previous work.19 It also quench bath temperatures were monitored by a type-K ther- ~ acted as a constraining, compressive force during cooling due mocouple (Table II). After 10 min of sitting in the quenchant to the significantly larger (30–60 9) linear coefficient of to cool, the ampoule was removed from the stainless steel holder, dried, and moved to the glovebox. The ingot, still inside the ampoule (SC) or ampoules (DC, see Figure 3(c)), (a) was then annealed at Tan (Table II) for 6.7 h at a heating rate of 2°C/min and a cooling rate of 1.5°C/min to help relieve the (b) thermal stresses inherent in the glass after quenching. Following annealing, the outer tubes of the DC specimens were removed (Figure 3(d)). For the SC ampoule, the ingot (c) was removed from the quartz tube, but for the DC specimens the ingot remained inside the inner ampoule and copper (d) sheath for subsequent processing. The SC ingot (CGA-1) was casted in resin and curing was assisted by a pressure chamber 5 (e) at ~3 9 10 Pa. The DC specimens were not mounted in resin as the sintered copper sheath around the inner ampoule pro- Fig. 3. Progression of double-containment method (all images vided enough support for further processing of the specimens. collected from CGA-3 with the “base” of the ampoule at the left and Specimens were cut with either a variable speed diamond the “top” of the ampoule at the right. (A) Stoichiometric quantities saw (Buehler, Lake Bluff, IL) or a CT400 diamond wire saw of constituents batched into a 10 9 12 mm (ID 9 OD) fused quartz (Diamond Wire Technology, LLC, Colorado Springs, CO) ampoule, evacuated, and sealed. (B) Inner ampoule is loaded into a into ~1–1.5-mm thick discs perpendicular to the longitudinal larger, 19 mm 9 22 mm, ampoule and copper powder is added to the void between the two ampoules, the outer ampoule is evacuated axis of the ingot. Specimens were polished with an LP-01 and sealed (prior to heat treatment). (C) After heat treatment. (D) Syntron vibratory polisher (FMC Technologies, Saltillo, MS) Inner ampoule with sintered copper jacket removed from outer to a <1 lm finish and then characterized. The processing ampoule. (E) Ingot removed from inner ampoule. The scale-bar is variables that were altered between experiments are presented valid for all photographs. in Table II. 8.18

2164 Journal of the American Ceramic Society—Riley et al. Vol. 95, No. 7

(2) Microscopy (3) X-ray Diffraction The various phases present in these specimens were so close in XRD was performed on powders of all the samples with a composition to the 1-1-2 crystalline material and residual Bruker® D8 Advance (Bruker AXS Inc., Madison, WI), amorphous matrix that they were not distinguishable with equipped with a CuKa target at 40 kV and 40 mA. The basic microscopy unless polarized light or a strong composi- instrument had a LynxEyeTM position-sensitive detector with tional contrasting method was implemented (such as scanning an angle range 3° 2h. Scan parameters used for sample analy- electron microscopy with backscattered electron contrast sis were 5–110° 2h with a step of 0.015° 2h and a 0.3-s dwell (SEM-BSE). Cross-sectioned specimens were observed under at each step. JADE 6© (Materials Data, Inc., Livermore, cross-polarized light at various magnifications (12.5–1000 9) CA), Bruker AXS DIFFRACplus EVA, and Bruker AXS To- with a Leitz Orthoplan optical microscope (Leica Microsys- pas 4.2 software were used to identify and quantitate phase tems GmbH, Wetzlar, Germany). Cross-polarized light optical assemblages. XRD analysis was performed on both polished microscopy provided the ability to distinguish between amor- discs and powdered specimens. phous, polycrystalline 1-1-2, and polycrystalline 3-2-4 phases. Laser confocal scanning microscopy was performed with a IV. Results Keyence VK9710 microscope (Keyence Corporation of Amer- ica, Elmwood Park, NJ) and a 50 9 objective to look for dif- Figures 4(b)–(f) provides a summary of the cross-sectioned ferences in optical appearance between the phases as well as to appearance of most of the samples discussed here and measure the variation in height across the specimen. Table III provides a summary of the estimated phase concen- Polished specimens were also examined with SEM-BSE. trations from each experiment. Chronologically, CGA-1 was Two different SEMs were used during analysis and included the first experiment we performed where the 3-2-4 phase was a JSM-5900 (JEOL, Ltd., Tokyo, Japan) and an FEI Helios observed (Figs. 4(a) and (b)); however, the quantity of this 600 Nanolab (FEI Company, Hillsboro, OR). Energy disper- phase was so small (Fig. 4(a)) that it was disregarded at first. sive spectroscopy (EDS) spot, area, and dot-mapping analy- We achieved a nearly amorphous sample with the SC experi- ses were performed on the unknown phase (3-2-4) with the ment (CGA-2). The second time the 3-2-4 phase was synthe- JEOL 5900 SEM with an EDAX Si-drifted EDS detector sized, in CGA-3, it was observed in very large quantities at (Apollo XL, 30 mm2) and EDAX Genesis 6.2 software ~30%, by volume. The primary difference between these two (AMETEK, Berwyn, PA). Portions of the 3-2-4 crystals from experiments was the ampoule processing method. Thus, the CGA-4e were removed with a focused ion beam (FIB) increase in the 3-2-4 phase volume fraction was attributed to attachment and an OmniProbe 250 lift-out accessory on the the DC ampoule processing method. Experiment CGA-2 was FEI SEM for transmission electron microscopy (TEM). quenched in liquid gallium and was almost completely amor- TEM was performed with a JEOL 2010F equipped with a phous (see Table III). Experiments CGA-3, CGA-4, and field emission gun and digital images were captured with a CGA-5 all had similar appearances though slight differences Gatan bottom-mounted charge-coupled device camera and in phase distribution. Experiment CGA-6 turned out very DigitalMicrographTM software (Gatan Inc., Pleasanton, CA). different where almost the entire ingot was the 3-2-4 phase. Specimens were analyzed in bright field, selected area, and Energy dispersive spectroscopy analysis was performed convergent beam electron diffraction modes. Chemical analy- with SEM and TEM and revealed a higher at.% of Cd in sis of selected locations was performed with EDS analysis the new, unknown phase than was present in the 1-1-2 phase with an Oxford Inca system (Oxford Instruments PLC, Tub- or the glass (Fig. 5). The EDS compositional data is pre- ney Woods, Abingdon, U.K.). sented in Table IV and an average simplified composition of

Fig. 4. Visual comparison of the CGA-# samples. Here, cross-polarized light optical micrographs (Leitz microscope, 12.5 9) are presented for each sample except for CGA-5, which appeared very similar to CGA-4, for comparison purposes. The 3-2-4 crystallites in (d) and (e) are so small that they do not appear at these magniﬁcations.

8.19

July 2012 Cd3 Ge2 As4: A New Crystalline Cd-Ge-As Phase 2165

Table III. Summary of Phase Quantities Observed in the cal ingot. However, the corner of a specimen prepared with a Various Experiments polished examination of a longitudinal and transverse face revealed that the 3-2-4 phase grew as platelets. Sample ID %1-1-2 %3-2-4 %Am. Other Figure 6 shows an example of the 1-1-2 and 3-2-4 crystal growth patterns along the perimeter of polished discs cut per- CGA-1 89 1 10 N/A pendicular to the longitudinal axis from specimens of CGA-4. CGA-2 5 0 95 N/A These projections were observed in several locations (i.e., 30– CGA-3 30 30 40 N/A 40) along the perimeters of these specimens and ~95% of the CGA-4 10 40 50 N/A 1-1-2 crystals were observed within ~1 mm of the perimeter CGA-5 5 50 45 N/A † as evidenced by the 1-1-2 polycrystallites. As the crystals CGA-6 <0.1 95 4 <1 have plate-like morphology in the bulk of the ingot and orig- Volume percentages are estimated from an average analysis of various spec- inate out of single nucleation sites as viewed in perpendicular imens from each sample with CPLM where %1-1-2, %3-2-4 and %Am denote specimen orientations, we believe that they are triangular- the percent of the 1-1-2, 3-2-4, and amorphous phases, respectively. shaped plates where the apex is on the outer perimeter sur- N/A, Not Applicable. † face of the cylindrical ingot. This covers other phases present that include amorphous and crystalline Confocal laser scanning microscopy, which is inherently phases of various compositions (see text). polarized, provided interesting contrast between the different phases present in the specimens (Fig. 6(b)), comparable to the the new phase is Cd3Ge2As4. The increased Cd content in cross-polarized light micrographs obtained (Fig. 6(a)). Confo- the new phase is evident from the BSE micrograph with cal laser scanning microscopy did not provide resolution atomic number contrast as well as in the elemental map that between different orientations of 1-1-2 crystallites as was also shows Cd depletion adjacent to the crystal (Fig. 5). The obtained with cross-polarized light microscopy, evidenced by dark streaks present in the core of the 3-2-4 crystals in Fig. 5 the color variations in the different grains. However, the con- were found to be amorphous with TEM selected area diffrac- focal laser scanning microscope was able to observe the topol- tion and had a composition matching closely to the composi- ogy of polished specimens. The topology maps revealed that tion of the amorphous regions as observed with SEM the 3-2-4 phase was lower than the surrounding amorphous (Table IV). phase, by an average of 120 ± 10 nm. Thus, we believe that Initial queries about the 3-2-4 crystals included the mecha- the 3-2-4 phase is slightly softer than the residual bulk amor- nism of nucleation and crystallization as well as the crystallo- phous glass and 1-1-2 phase, and was preferentially removed graphic morphology. Initial observations with cross-polarized during the polishing process. light microscopy led us to believe that these new crystals When cross-sections were made of CGA-4 and CGA-5 per- were rods protruding from the perimeter of the specimen and pendicular to the longitudinal axis of the ingot, the crystals growing inward toward the longitudinal axis of the cylindri- were optically observed in the two distinct widths of narrow

(a) (b) (c) (d)

Fig. 5. (A) CGA-4e sample location where an SEM (JEOL 5900) dot map was performed (region denoted by “DM”) and the location where samples were removed with the FIB (labeled “B” and “C”) for TEM analysis that correspond to (B) and (C). (B) FIB section micrograph (region “B” from (A)) captured with high angle annular dark ﬁeld (HAADF) in scanning transmission (STEM) mode where the 3-2-4 phase appears darker than the amorphous background. FIB section micrograph (region “C” from (A)) captured with an Everhardt-Thornley secondary electron detector where the 3-2-4 phase appears brighter than the amorphous background. (D) TEM micrograph showing regions of EDS analysis on (C) where the specimen is ﬂipped 180°C from the micrograph (C). Along the bottom row is a Backscattered SEM dot map from in the region outlined by the “DM” box in (A) showing the elemental distribution of Cd, Ge, and As.

Table IV. Compositional Information (in at.%) via EDS for the Diﬀerent Regions Observed in These Specimens: the CdGeAs2 or 1-1-2 Phase, the Unknown Phase (~Cd3Ge2As4), and the Amorphous Matrix

SEM TEM Collection Method: Element Cd Ge As Cd Ge As

Unknown phase (3-2-4) 32.75 (0.68) 21.65 (0.26) 45.60 (0.42) 33.89 (0.29) 22.35 (0.20) 43.76 (0.08) 1-1-2 25.05 (0.44) 24.38 (0.23) 50.57 (0.41) N/A N/A N/A Amorphous matrix 20.81 (2.87) 25.14 (0.51) 54.05 (2.42) 17.29 30.71 52.00 These values were quantiﬁed using the Cd-La, Ge-Ka, and As-Ka lines with EDS from both SEM and TEM. SEM values were collected on CGA-3 and TEM values were collected on specimen CGA-4 (Figure 5). Calculated error based on standard deviation is listed in parenthesis. N/A, Not Applicable.

8.20

2166 Journal of the American Ceramic Society—Riley et al. Vol. 95, No. 7

(a) (b)

Fig. 6. Growth patterns of the 3-2-4 phase from nucleation sites along the perimeter of the ingot inward toward the longitudinal axis as observed on cross-sections prepared perpendicular to the longitudinal axis of the sample. Here, these patterns are observed with cross-polarized light on an optical microscope (a) and with a confocal laser scanning microscope (b) on polished samples. The 1-1-2 phase shows up as purple- red (a) or dark gray (b) clusters of polycrystallites and the amorphous regions make up the remainder of the micrograph.

(Fig. 5(a), “C”) and wide (Fig. 5(a), “B”). It appeared as though the crystals with different widths were captured at different growth directions or angles, approximately perpendicular to one another. In close observation of the sections removed with the FIB (Figs. 5(b) and (c)), we determined that the 3-2-4 crystals were thin sheets (~5–10 lm) laminated by amorphous Cd-Ge-As material (200–400 nm thick), as verified with TEM, with similar composition as the bulk (Table IV). For CGA-4 and CGA-5, the widths of the 3-2-4 crystallites were measured in cross-polarized light micrographs captured in a few different regions on one of the specimens. The thickness of the 3-2-4 plates (in the narrow orientation) was ~10 ± 3 lm near the perimeter of the specimen and 25 ± 5 lm near the center of the specimen. The width in the wide orientation was 40 ± 5 lm near perimeter and ~100 ± 10 lm near center; thus, the 3-2-4 crystallites grew in width as they approach the center of the ingot. The plates protruded from the perimeter toward the center of the ingot at lengths of 5 mm (Fig. 6). A lower density of 3-2-4 crystallites was found toward the center of the ingot than the perimeter. This is consistent with the fact that the 3-2-4 crystals require a large amount of Cd to satisfy the crystal structure and that the crystals increased in size toward the center of the ingots. The purpose of CGA-6 was to better understand the chemistry limitations on 3-2-4 crystal formation because it was not batched Cd-deficient as opposed to all of the other experiments. For CGA-6, nearly the entire ingot was found to be crystalline (see Table III). Investigations with cross- polarized light microscopy (Fig. 4(f)) and XRD (Fig. 7) revealed that the vast majority of the crystalline phase was the 3-2-4 phase (see Table III). For CGA-6, the crystallization process appeared similar to θ CGA-3, CGA-4, and CGA-5, with small crystallites on the perimeter and separate dendritic grains growing toward the Fig. 7. Compilation of diffraction patterns from various samples core, but the phases observed were different from those showing the progression of amorphous (CGA-2) to 100% 1-1-2 (CGA-1), to mixed 1-1-2 and 3-2-4 (CGA-3), and then to nearly observed in the previous samples. Instead of an amorphous 100% 3-2-4 (CGA-6). Also included in the plot are a calculated hkl rim as observed in CGA-3, CGA-4, and CGA-5, the rim of pattern for the 3-2-4 phase (hkl phase (calc)), a diffraction pattern CGA-6 was composed of the following phases: for 1-1-2 phase from the ICSD database (16736),42 the calculated pattern for CGA-3 including the 1-1-2 phase and the hkl pattern for 1. 3-2-4 phase at the perimeter of the specimen; the 3-2-4 phase (calc) compared with the measured pattern for CGA- 2. very small quantities of a Cd-As phase, with a small 3 (meas), and the difference pattern (Diff. Spec.) for the calculated fraction of Ge as measured with SEM-EDS, another versus measured data. new Cd-Ge-As phase with crystalline-like morphology, growing in the interstitial regions of 3-2-4 grains a = b = 7.735 A˚ , c = 27.40 A˚ , a = b = 90°, and c = 120°. (~5–10 lm in size); From this data, an hkl fit was performed on the powder dif- 3. amorphous regions deficient in Cd from the as-batched fraction pattern of CGA-3 with Topas 4.2 software. The cal- 3-2-4 composition; and culated diffraction pattern was generated including the 4. a very small quantity of 1-1-2 material (≲ 0.1 vol%) contribution from the known 1-1-2 phase.42 observed with cross-polarized light microscopy. The XRD results presented in Fig. 7 demonstrate that the Preliminary single-crystal diffraction analysis (not shown calculated hkl pattern for the 3-2-4 phase fits well to the dif- here) on the 3-2-4 crystals determined that the crystals had ferent specimens and the differences in the fit are minimal. hexagonal lattice constants, a trigonal system, and a trigonal Figure 7 shows a progression of crystallization from the space group of P-3. The unit cell constants are thought to be amorphous sample (CGA-2), to the sample completely

8.21

July 2012 Cd3 Ge2 As4: A New Crystalline Cd-Ge-As Phase 2167 crystallized with the 1-1-2 phase (CGA-1), to the sample The thermal compression of the sintered copper jacket was almost completely crystallized with the 3-2-4 phase (CGA-6). evidenced by the fact that the copper sheath pulled away The calculated hkl pattern shows a good fit to the measured from the outer ampoule and compressed very tightly on the pattern for CGA-3 with minor differences. The CdAs-rich inner ampoule – so tightly that it could not be removed regions observed in CGA-6 were so small that they were not unless it was cut off. It is worth noting that the porosity of observed in the diffraction pattern (Table III). the sintered copper sheath was rather high at ~ 46 ± 2.8 vol %, as measured by optical microscopy. The increase in thermal mass of the DC method with the additional quartz tube V. Discussion and copper powder (mCu > ms) is another key difference The 3-2-4 phase was created with the DC ampoule technique between the SC and DC processing methods. The additional several times in separate experiments under different condi- mass resulted in a slower cooling rate, allowing more time tions, and is therefore a reproducible phase. The 3-2-4 con- for the nuclei present in the melt to undergo crystallization. tent in specimens batched with a 1-1-2 bulk composition was It remains uncertain which of the two crystalline phases observed to increase as the Tp was increased and the Tan was (1-1-2 and 3-2-4) grew first; they grew in conjunction with decreased (Table II). The yield of the amorphous phase one another along the perimeter, where the cooling rate of remained relatively consistent in CGA-3, CGA-4, and the sample and the population density for both phases CGA-5. where at their maxima. For CGA-3, CGA-4, and CGA-5, it We believe that the nucleation and growth of this new 3- appeared that the three phases present were formed in the 2-4 phase was caused by one of the following: (1) the pres- following order based on the direction of crystal growth: (1) ence of a chemical impurity in these specific reactants, (2) the the very outer rim was quenched into the amorphous state specific cleaning protocol used in these experiments (that (fastest cooling rate), (2) very small 1-1-2 crystallites formed might create special heterogeneous nucleation sites), (3) the (slower cooling rate), and (3) the 3-2-4 crystallites grew off DC processing method used in this work, and/or (4) some- of 1-1-2 crystals (slowest cooling rate). As supercooling is thing else not considered. The likelihood of a chemical impu- possible in Cd-Ge-As materials,22 the temperature of the rity being the cause of the growth of 3-2-4 crystals, that is, a sample is thought to have been below the Tc for the 1-1-2 nucleating agent, is low due to the high purity of the reac- phase at the point of sample crystallization (438°C–450°C 19,23 tants ( 6N’s). However, this possibility was not ruled out. for CdGeAs2). In CGA-3, CGA-4, and CGA-5, the The! ampoule cleaning technique could potentially aid in composition of the interstitial amorphous phase between the the nucleation of the 1-1-2 or 3-2-4 crystals on the inner sur- 3-2-4 grains did not noticeably change radially from the 3-2- faces of the fused quartz ampoule. The acid-etching step was 4. If the 3-2-4 grains propagated within the sample toward shown to increase the surface area of the quartz where the the center of the ingot through solid-state diffusion, the roughness number, Ra, measured using atomic force micros- composition of the interstitial amorphous phase would copy was 3.9 9 higher in the acid etched fused quartz versus change with Cd-deficient regions near the 3-2-4 crystals and washing the quartz with DIW only.35 However, considering a higher Cd concentration further removed from the 3-2-4 that the 3-2-4 phase was present in CGA-1, where a carbon crystals. coating was applied to the inner surface of the ampoule, and The Tg and Tm of the residual glass phase present in these the 3-2-4 phase was not present in CGA-2, where the specimens after crystallization are presumed to have changed ampoule was quenched in liquid gallium, the inner ampoule as the composition changed from the as-batched composition 23,32,46,47 surface preparation is an unlikely cause of the 3-2-4 appear- of 1-1-2 where Tm ~ 665°C–670°C and Tg ~ 390°C. ance. Nevertheless, the inner ampoule surface preparation Both values are expected to have increased as x decreased in likely helped nucleate crystallization in the experiments with CdxGeAs2 considering that the compound GeAs2 (x = 0, or 20 a lower quenching rate (DC experiments). 0-1-2) has Tm = 732 °C. However, we are not aware of any II IV V For CGA-2, the liquid gallium quenchant effectively wet- glass-formation work done in II-IV-V2 (A B C2 ) materials ted the fused quartz ampoule, especially when compared with deficient in the Group II component. A Cd-deficient Cd-Ge- water at these quenching temperatures (T ~ 300°C–800°C) As glass was found in all of the crystallized specimens. Thus, where the water quenchant was above the boiling tempera- CdxGeAs2 glasses where x < 1 might prove to be good glass- ture (T > 100°C). Therefore, a layer of steam formed forming compounds that warrant further investigation. between the outer ampoule wall and the quenchant at the According to the observations that were made for CGA- moment of quenching. Also, the thermal conductivity (k in 3, CGA-4, and CGA-5, the composition of the amorphous 1 W·(m·K)À ) of gallium is rather high at kGa ~ 28.65 (77°C)– phase is 18–20 at.% Cd with the balance being GeAs2 38.27 (277°C), where the k for liquid water is low at (Table IV). For CGA-6, the phases were dissimilar to those k ~ 0.56 (1.9°C)–0.67 (97°C) and the k for steam is in the other samples and this was due to the different start- H2O;liq: 43–45 even lower at kH2O;gas: ~ 0.026 (127°C)–0.041 (277°C). ing composition of Cd33.3Ge22.2As44.4 (3-2-4) versus Thus, the gallium was expected to pull the heat out of the Cd25Ge25As50 (1-1-2). As a Cd-rich CdGeAs2 phase is not ampoule faster than the water quenchant, providing a higher stable in the amorphous state, the residual phases present cooling rate. The increased wettability on fused quartz com- after the 3-2-4 crystalline material was removed from the bined with the higher thermal conductivity of gallium matrix were (1) a slightly As-deficient, Cd-doped GeAs2 increased the quenching rate of CGA-2. This allowed for su- amorphous phase, (2) an As-rich phase with composition percooling past Tc, allowing for an almost completely amor- close to Cd2Ge3As11.3 (Cd13Ge19As68), and (3) the CdAs- phous ingot, which was not the case in any of the other rich phase. The CdAs-rich phase was observed along the samples. grain boundaries of the 3-2-4 crystals along the perimeter The largest quantities of 3-2-4 crystals were observed in and in the bulk of the specimen in polished cross-sections. specimens processed with the DC method, that is, the method with the slowest cooling rates of those tested. In our VI. Conclusions previous work,19 we discussed the benefits of the DC method and one of those benefits was that the thermal compression A new Cd-Ge-As phase was made with a double-contain- of the sintered copper jacked compressed the inner quartz ment ampoule method and the measured composition was tube, thereby compressing the material inside of the inner Cd3Ge2As4 (3-2-4). These crystals were made in large quanti- ampoule. The density of amorphous Cd-Ge-As is higher than ties through several similar experiments. These 3-2-4 crystals that of the crystalline analogs,19,38 thus this thermal compres- grow with a plate-like morphology and are presumed to sion might have forced the materials into the amorphous nucleate from 1-1-2 crystals that appear first on the perimeter state through the quenching process between T and T . of the cylindrical ingot during the cooling process. m c 8.22

2168 Journal of the American Ceramic Society—Riley et al. Vol. 95, No. 7

More work is needed to fully understand why these crystals 17G. W. Iseler, H. Kildal, and N. Menyuk, “Optical and Electrical Proper- form the structures presented here. At the current time the ties of CdGeAs2,” J. Electron. Mater., 7 [6] 737–55 (1978). 18B. R. Johnson, B. J. Riley, J. V. Crum, S. K. Sundaram, C. H. Henager growth of the 3-2-4 crystals is thought to be facilitated by the Jr, D. L. Edwards, and A. L. Schemer-Kohrn, “Discovery of Metastable DC quenching method that provides a specific cooling rate Crystalline Phases in Cd-Ge-As Glass,” J. Am. Ceram. Soc., 91 [2] Cover and a compressive force during the cooling process, forcing the (2008). 19 solidified Cd-Ge-As ingot into a higher density state. B. R. Johnson, B. J. Riley, S. K. Sundaram, J. V. Crum, C. H. Henager Jr, Y. Zhang, V. Shutthanandan, C. E. Seifert, R. M. Van Ginhoven, C. E. These materials have several potential applications as Chamberlin, A. A. Rockett, D. N. Hebert, and A. R. Aquino, “Synthesis and those identified for similar Cd-Ge-As compounds although Characterization of Bulk, Vitreous Cadmium Germanium Arsenide,” J. Am. measurements to validate these possibilities have yet to be Ceram. Soc., 92 [6] 1236–43 (2009). 20 realized due to the small size of the 3-2-4 crystallites. These M. Hansen and K. Anderko, Constitution of Binary Alloys. 2nd ed, pp. 1305. McGraw-Hill Book Company, Inc., New York, 1958. applications include photoconductivity, semiconductivity, 21A. S. Borshchevskii, N. A. Goryunova, F. P. Kesamanly, and D. N. Nas- II IV V and nonlinear optics. ledov, “Semiconducting A B C2 Compounds,” Phys. Status Solidi B, 21 [1] 9–55 (1967). 22A. S. Borshchevskii and N. D. Roenkov, “Phase Diagram of the Cadmium- Acknowledgments Germanium-Arsenic System,” Russ. J. Inorg. Chem., 14 [8] 1183–6 (1969). 23K. S. Hong, Y. Berta, and R. F. Speyer, “Glass-Crystal Transition in II-IV- Pacific Northwest National Laboratory (PNNL) is operated for the U.S. V2 Semiconducting Compounds,” J. Am. Ceram. Soc., 73 [5] 1351–9 (1990). 24 Department of Energy by Battelle under Contract DE-AC05-76RL01830. J. C. Mikkelsen Jr and H. Y.-P. Hong, “Cd2GeAs4: A New Ternary Phase This work was performed under contract with the Department of Energy’s in the Cd-Ge-As System,” Mat. Res. Bull., 9 [9] 1209–18 (1974). Nonproliferation and Verification R&D (NA-22) Program. Authors thank 25B. R. Pamplin and R. S. Feigelson, “New Phases in the Cd-Ge-As Sys- Clyde Chamberlin for his extensive work with specimen cutting and polish- tem,” Mat. Res. Bull., 14 [2] 263–6 (1979). ing. Authors also thank Carolyn Burns for running particle size analysis on 26K. T. Zawilski, P. G. Schunemann, and T. M. Pollak, “Glass Formation the copper powder used in the DC experiments and Angus A. Rockett and and Optical Properties of CdGeAs2 Alloys,” J. Cryst. Growth, 310 [7–9] 1897– Damon N. Hebert at the Materials Research Laboratory, located at the 903 (2008). University of Illinois at Urbana-Champaign for helpful discussions and assis- 27P. G. Schunemann and T. M. Pollak, “Ultralow Gradient HGF-Grown tance in characterization. A portion of this research was performed using ZnGeP2 and CdGeAs2 and Their Optical Properties,” Mater. Res. Bull., 23 [7] the Environmental Molecular Sciences Laboratory, a national scientific user 23–7 (1998). facility sponsored by the Department of Energy’s Office of Biological 28J. L. Shay and J. H. Wernick, Ternary Chalcopyrite Semiconductors: and Environmental Research and located at Pacific Northwest National Growth, Electronic Properties, and Applications. Pergamon Press, New York, Laboratory. 1975. 29 H. Kildal, “CdGeAs2 and CdGeP2 Chalcopyrite Materials for Infrared Nonlinear Optics”; Ph.D. dissertation Thesis, Stanford University, Stanford, References CA,1972. 30K. Nagashio, A. Watcharapasorn, K. T. Zawilski, R. C. DeMattei, R. S. 1 S. Sharma, K. S. Hong, and R. F. Speyer, “Glass Formation in Chalcopy- Feigelson, L. Bai, N. C. Giles, L. E. Halliburton, and P. G. Schunemann, rite Structured Semiconducting Compounds,” J. Mater. Sci. Lett., 8 [8] 950–4 “Correlation Between Dislocation Etch Pits and Optical Absorption in (1989). CdGeAs2,” J. Cryst. Growth, 269 [2–4] 195–206 (2004). 2 K. T. Stevens, L. E. Halliburton, S. D. Setzler, P. G. Schunemann, and T. 31S. M. Evans, N. Y. Garces, R. C. DeMattei, R. S. Feigelson, K. T. Zawil- M. Pollak, “Electron Paramagnetic and Electron-Nuclear Double Resonance ski, N. C. Giles, and L. E. Halliburton, “Electron Paramagnetic Resonance 2+ Study of the Neutral Copper Acceptor in ZnGeP2 Crystals,” J. Phys. Condens. and Electron-Nuclear Double Resonance Study of Mn Ions in CdGeAs2 Matter, 15 [10] 1625–33 (2003). Crystals,” Phys. Status Solidi B, 243 [15] 4070–9 (2006). 3 U. N. Roy, M. Groza, Y. Cui, A. Burger, Z. W. Bell, and D. A. Carpenter, 32P. Leroux-Hugon, “Properties of Some Ternary Semiconductor Com- “Crystal Growth, Characterization and Fabrication of AgGaSe2 Crystals as a pounds,” C.R. Hebd. Seances Acad. Sci. (France), 256 [1] 118–20 (1963). Novel, Material for Room-Temperature Radiation Detectors”; pp. 177–80 in 33B. R. Pamplin, T. Kiyosawa, and K. Masumoto, “Ternary Chalcopyrite Proc. of the SPIE, Vol. 5540. Edited by A. Burger, R. B. James, and L. A. Compounds,” Prog. Crystal Growth Charact., 1 [4] 331–87 (1979). Franks. The International Society for Optical Engineering, Bellingham, WA, 34G. Lianxing, Z. Bing, and Z. Wenlan, “Chalcopyrite Intergrowths in 2004. Sphalerite in the Meixian Lead-Zinc Deposit, Fujian Province and Their Met- 4 S. Cho, S. Choi, G. Cha, S. C. Hong, Y. Kim, Y. Zhao, A. J. Freeman, J. allogenic Significance,” Chin. J. Geochem., 17 [4] 311–9 (1998). B. Ketterson, B. J. Kim, Y. C. Kim, and B. Choi, “Room-Temperature Ferro- 35B. J. Riley, B. R. Johnson, S. K. Sundaram, M. H. Engelhard, R. E. magnetism in (Zn1-xMnx)GeP2 Semiconductors,” Phys. Rev. Lett., 88 [25] Williford, and J. D. Olmstead, “Pressure–Temperature Dependence of Nano- 257203, 4pp (2002). wire Formation in the Arsenic–Sulfur System,” Phys. Chem. Glass, 47 [6] 675– 5 B. R. Johnson, J. V. Crum, S. K. Sundaram, R. M. Van Ginhoven, C. E. 80 (2006). Seifert, B. J. Riley, and J. V. Ryan, “DC Ionization Conductivity of Amor- 36A. A. Vaipolin, F. M. Gashimzade, N. A. Goryunova, F. P. Kesamanly, phous Semiconductors for Radiation Detection Applications,” IEEE Trans. D. N. Nasledov, E. O. Osmanov, and Y. V. Rud, “Investigation of the Physi- Nucl. Sci., 56 [3] 863–8 (2009). cochemical and Electric Properties of Crystals of Some Ternary Semiconductor 6 II IV V G. A. Medvedkin, K. Hirose, T. Ishibashi, T. Nishi, V. G. Voevodin, and Compounds of the A B C2 Type,” Izv. Akad. Nauk SSSR Ser. Fiz. K. Sato, “New Magnetic Materials in ZnGeP2-Mn Chalcopyrite System,” (USSR), 28 [6] 1085–9 (1964). J. Cryst. Growth, 236 [4] 609–12 (2002). 37N. A. Goryunova, F. P. Kesamanly, and E. O. Osmanov, “Preparation 7 Y. Zhao, S. Picozzi, A. Continenza, W. T. Geng, and A. J. Freeman, “Pos- and Some Properties of CdGeAs2 Single Crystals,” Fiz. Tverd. Tela (USSR), sible Impurity-Induced Ferromagnetism in II-Ge-V2 Chalcopyrite Semiconduc- 5 [7] 2031–2 (1963). tors,” Phys. Rev. B Condens. Matter, 65 [9] 094411, 6pp (2002). 38H. Hruby, “Evaluation of Glass-Forming Tendency by Means of DTA,” 8 R. L. Byer, H. Kildal, and R. S. Feigelson, “CdGeAs2 - A New Nonlinear Czech J. Phys., B22 [11] 1187–93 (1972). Crystal Phasematchable at 10.6 lm,” Appl. Phys. Lett., 19 [7] 237–40 (1971). 39T. A. Hahn, “Standards, Copper, Methods, Thermal Expansion of Copper 9 G. D. Boyd, E. Buehler, F. G. Storz, and J. H. Wernick, “Linear and Non- from 20 to 800 K–Standard Reference Material 736,” J. Appl. Phys., 41 [13] II IV V linear Optical Properties of Ternary A B C2 Chalcopyrite Semiconductors,” 5096–101 (1970). IEEE J. Quantum Electron., QE-8 [4] 419–26 (1972). 40SRM-736, Copper-Thermal Expansion. National Institution of Standards 10 R. L. Byer, “Optical Parametric Oscillators”; pp. 587–702 in Quantum and Technology, Gaithersburg, MD, 1990. Electronics: A Treatise, Edited by H. Rabin and C. L. Tang. Academic Press, 41SRM-739: Fused-Silica Thermal Expansion. National Institute of Stan- New York, 1977. dards and Technology, Gaithersburg, MD, 1991. 11 P. G. Schunemann and T. M. Pollak, “Single Crystal Growth of Large, 42H. Pfister, “Kristallstruktur von terna¨ ren Verbindungen der Art II IV V Crack-Free CdGeAs2,” J. Cryst. Growth, 174 [1–4] 272–7 (1997). A B C2 ,” Acta Cryst., 11, 221–4 (1958). 12 H. Katzman, T. Donohue, W. F. Libby, H. L. Luo, and J. G. Huber, “A 43M. J. Duggin, “The Thermal Conductivity of Liquid Gallium,” Phys. High-Pressure Superconducting Polymorph of Cadmium Tin Diarsenide,” Lett., 29A [8] 470–1 (1969). J. Phys. Chem. Solids, 30 [6] 1609–11 (1969). 44M. L. V. Ramires, C. A. Nieto de Castro, Y. Nagasaka, A. Nagashi- 13 K. L. Vodopyanov and P. G. Schunemann, “Efficient Difference-Fre- ma, J. Assael, and W. A. Wakeham, “Standard Reference Data for the quency Generation of 720-lm Radiation in CdGeAs2,” Opt. Lett., 23 [14] Thermal Conductivity of Water,” J. Phys. Chem. Ref. Data, 24 [3] 1377–81 1096–8 (1998). (1995). 14 V. Rud, Y. Rud, I. Polushina, T. Ushakova, and S. Iida, “Observation of 45R. W. Powell, C. Y. Ho, and P. E. Liley, “Thermal Conductivity of Record Electron Hall Mobility in CdGeAs2 Single Crystals,” Jpn. J. Appl. Selected Materials”; 165 pp. in National Standard Reference Data Series- Phys., 39, 266–7 (2000). National Bureau of Standards, Category 5-Thermodynamic and Transport 15 T. S. Moss, “A Relationship Between the Refractive Index and the Infra- Properties, Vol. 8. U.S. Department of Commerce, Washington, D.C., 1966. Red Threshold of Sensitivity for Photoconductors,” Proc. Phys. Soc. B, 63 [3] 46M. J. Harrison, A. P. Graebner, W. J. McNeil, and D. S. McGregor, 167–76 (1950). “Carbon Coating of Fused Silica Ampoules,” J. Cryst. Growth, 290 [2] 597– 16 H. Katzman, T. Donohue, W. F. Libby, and H. L. Luo, “A High-Pressure 601 (2006). 47 Superconducting Polymorph of Cadmium Germanium Diarsenide,” J. Phys. S. H. Risbud, “Processing and Properties of Some II-IV-V2 Amorphous Chem. Solids, 30 [12] 2794–5 (1969). and Crystalline Semiconductors,” Appl. Phys. A, 62 [6] 519–23 (1996). h

8.23

Manuscript submitted to Journal American Ceramic Society, November 2012

Kelvin probe force microscopy of metal contacts on amorphous cadmium germanium arsenide

Damon N. Hebert, Allen J. Hall, Angel R. Aquino, Angus A. Rockett University of Illinois, Materials Science & Engineering Department Urbana, IL

Richard T. Haasch University of Illinois, Frederick Seitz Materials Research Laboratory Urbana, IL

Bradley R. Johnson, Brian J. Riley Pacific Northwest National Laboratory, Glass Science & Processing Richland, WA

8.1 ABSTRACT

Au and Al/Mg bilayers have been studied as possible contacts to amorphous CdGe0.65As2

(CGA) semiconductors. Kelvin probe force microscopy (KPFM) and ultraviolet photoelectron spectroscopy (UPS) were used to measure the contact potential difference between these metals and an amorphous bulk CGA sample. Au was found to have a contact potential difference of -

500 meV relative to the CGA by KPFM and -780 meV by UPS. The bilayer contact with Al on

Mg on CGA showed a 400 meV contact potential difference from the Al to the CGA based on

KPFM. The contact of Mg to CGA should be about 500 meV higher due to the contact potential difference between Mg and Al. The contact potential difference for Mg/CGA estimated from

UPS was 700 meV. Electrical measurements showed that the conductivity of our CGA exhibits a thermally-activated behavior with an activation energy of 520 meV. The bilayer contact should have yielded a Schottky type contact for both the measured and anticipated contact potential difference, although the temperature-dependent current/voltage curves showed the

8.24

behavior to be ohmic for all temperatures studied. Other contact metals were also tested and they did not produced effective Schottky contacts either, possibly due to interfacial reactions or second phases not detected in the CGA. However, the range of metals tested, the absence of second phases in the XRD, the low conductivity of the CGA and the consistent electric field observed from the contacts to the CGA with high resolution of the KPFM measurements tends to suggest that the contact was a Schottky barrier over most of its area but that some form of undetectable region caused a low-resistance contact.

8.2 1. Introduction

In crystalline form, the CdGexAs2 (CGA) semiconducting chalcopyrite compound and related materials are of interest in energy conversion applications and as infrared nonlinear optical materials.1,2 Novel processing approaches provide the possibility of amorphous-

3 crystalline p-n junctions. These compounds, as well as other similar II-IV-V2 chalcopyrite materials, are known to exhibit behaviors useful for nonlinear optics, switching, radiation detection and photovoltaic device applications.4-6 For radiation detection purposes, a reproducible low-leakage Schottky barrier contact is desirable because under high bias fields, the device dark current can exceed the signal generated by low-levels of ionizing radiation. For example, it is often necessary to be able to detect nA-level signals at bias fields of ~1000 V/cm.

Therefore, fabrication of reproducible Schottky contacts to amorphous CGA is required for the material to be employed as the active sensing material in a radiation detection device.

8.25

This work reports on the fabrication of two types of metal/CGA contacts and their analysis with contact potential differences as measured by Kelvin probe force microscopy

(KPFM) and work functions as measured by ultraviolet photoelectron spectroscopy (UPS).

KPFM is a non-contact atomic force microscopy (AFM) method that measures the contact potential difference (CPD) between the surface of the sample and the conductive AFM tip resulting from changes in the vacuum level. This is equivalent to the difference between the work function of the AFM tip (φtip) and the work function of the specimen surface (φsc), CPD =

(φtip - φsc)/q, where q is the elementary charge. KPFM has been used, for example, to study the surface potential of chalcopyrite thin films7 and lateral Ni/AlGaN Schottky junctions8 as well as to observe surface charging at the edge of metal/GaN Schottky contacts.9 In this study, contact behavior was also analyzed with temperature dependent current-voltage (I-V) measurements with three different contact configurations. I-V and capacitance-voltage (C-V) curves have previously been used to characterize Al-based Schottky contacts on chalcopyrite I-III-VI2 semiconductors,10 but never before on amorphous CGA compounds.

For a semiconductor, the work function is just the sum of the electron affinity, χsc, and the separation between the Fermi energy level (EF) and the conduction band (Ec), Ea = Ec-EF.

Therefore the work function of a semiconductor depends on its doping and the position of the

Fermi level. This study reports work functions of electron-beam evaporated Au films and of the

CGA sample itself characterized with UPS. Finally, X-ray photoelectron spectroscopy (XPS) was used to determine the position of the Fermi level.

In preliminary (unpublished) experiments, several metals were studied as possible CGA contacts with varying work functions (φ). Au (φ = 5.10-5.47 eV), Pt (φ = 5.65-5.70 eV), Ni (φ =

5.04-5.35 eV), Ti (φ = 4.33 eV), Ag (φ = 4.26-4.74 eV), Mg (φ = 3.66 eV), and Al (φ = 4.06-

8.26

4.41 eV) 11 were all used. Although diode-like behavior was observed in one particular case (Mg

– CGA – Mg contact configuration), the lack of reproducibility called for a more organized study and was the motivation for the current work.

Au and Al/Mg were chosen as candidate contacts for KPFM analysis because Au has a high work function (5.10-5.47 eV) and Mg has a low work function (3.66 eV). Al was used as a capping metal on the Mg to prevent post-deposition oxidation and because Al also has a relatively low work function (4.06-4.41 eV).11 Thus, based on the expected work function of

CGA and its small mobility gap (0.65-1.1 eV),12 one metal contact should have produced a

Schottky barrier and the other an ohmic contact regardless of the (intrinsic) doping type of the

CGA.

8.3 2. Experiment

2.1 Material and Contact Fabrication

CGA glass samples of the composition CdGe0.65As2 were synthesized by water quenching molten liquids enclosed in double containment ampoules to form bulk, vitreous, crack-free ingots. Details of the fabrication process and subsequent characterization are published elsewhere.12 As explained in Reference 12, CGA is a fragile glass forming system.

Consequently, candidate samples selected for KPFM analysis in this study were screened by cross-polarized reflected light microscopy and X-ray diffraction to ensure that they were uniformly amorphous. For the sample chosen for study by KPFM analysis, X-ray diffraction showed it to be amorphous, although pair correlation function analysis revealed a high degree of short range order out to the 4th nearest neighbor.12 The specimen had a bulk composition of

Cd1.00Ge0.65As2.00. Compositional analysis performed on similar specimens with particle induced

8.27

X-ray emission (PIXE), Rutherford backscattering spectroscopy (RBS), and secondary ion mass spectroscopy (SIMS) confirmed that the actual stoichiometry was uniform on scales measured by

SIMS and matched the intended “as-batched” stoichiometry.12 The sample mounted for KPFM was polished in a vibratory polisher with 1-µm diamond polish and cleaned in an ultrasonic cleaner. It was polished a second time with colloidal a silica polish solution. All specimens were cleaned with acetone, isopropanol and de-ionized water and dried with nitrogen gas immediately before contact deposition and before KPFM analysis.

The Au/CGA contacts and Al/Mg/CGA bilayer contacts were deposited with a Temescal electron-beam evaporation system on several CGA samples. Typical contact thicknesses were

60-80 nm, with the Al/Mg bilayer thicknesses approximately 25±5 nm Al on top of 45±5 nm of

Mg, with deposition rates of 0.05-0.15 nm/s. Contacts were approximately 0.8 mm in diameter.

For I-V curve measurements, contacts were spaced in a Van der Pauw geometry, 3-10 mm apart, around the periphery of ~10 mm diameter, disc-shaped bulk CGA substrates. However, for

KPFM analysis, contact edges were required to be more abrupt than the 0.8 mm diameter contacts used for I-V curve analysis. Therefore, patterned Au/CGA and Al/Mg/CGA contacts were deposited with transmission electron microscopy grids (SPI 2030N; 3 mm, 300 mesh, Ni) as shadow masks on the chosen CGA specimen for KPFM analysis. The masking procedure led to high contrast regions of metal on the CGA spaced by 25 µm (Figure 1).

2.2 Ultraviolet photoelectron spectroscopy (UPS) and X-ray photoelectron spectroscopy (XPS)

Surface work functions were measured with in situ UPS and spectra were acquired by means of a Physical Electronics PHI 5400 photoelectron spectrometer utilizing He I (21.2 eV) ultraviolet radiation as the excitation source. The UPS source was incident at an angle of 50°

8.28

while the electron detector had an acceptance angle of 20°. The UPS data were collected at an emission angle of 0° relative to the surface normal. Prior to UPS measurements, the Au and

CGA samples were sputtered for 30 seconds with 3 keV Ar+ rastered over a 3×3 mm2 (10

µA/cm2) area to remove surface contamination. Alignment of the Fermi level with the valence band edge (and thus doping type) of the CGA sample was examined with a Kratos Axis ULTRA high-performance X-ray photoelectron spectrometer with a monochromatic Al Kα (1486.7 eV) excitation source. For these techniques, spectral information was collected from a depth of 2-20 atomic layers, depending on the photoelectron escape depth (material dependent).

2.3 Kelvin probe force microscopy

An Asylum Research MFP3D AFM was used to measure surface topography (AC mode) and contact potential difference (KPFM, Nap mode), simultaneously, with a BudgetSensors

ElectriTap 300 Pt/Cr (25/5 nm) monolithic Si probe tip (BudgetSensors, Bulgaria). Samples were measured in ambient air conditions. Scan speeds of approximately 11 µm/s were used to ensure good tip-surface tracking and no bias voltage was applied to the tip during Nap-mode

(KPFM) scans. Data were aquired and analyzed with Asylum MFP3D software in IGOR Pro

(Asylum Research, Santa Barbara, CA, USA).

8.3.1 2.4 Electrical Measurements

Current-voltage curves were measured as a function of sample temperature with the same contact materials (Au and Al/Mg) as were measured in the KPFM scans. Four-point probe measurements on the CGA samples were performed with linearly-aligned contacts spaced 2 mm apart. Samples were housed in a sealed, evacuated sample chamber and mounted with Apiezon

L grease (M&I Materials Ltd., Manchester, UK) to a cold stage. The cold stage was cooled

8.29

using a CTI Cryogenics 8200 compressor with a temperature range of 40 to 300 K. A Lakeshore

330 autotuning temperature controller (Lakeshore Cryotronics, Inc., Westerville, OH, USA) with a TG-120-PL GaAlAs temperature sensor and a resistive heating element in the cold stage were used to maintain the specified sample temperature. An MMR H-50 Hall/Van der Pauw controller (MMR Technologies, Mountain View, CA, USA) was used to apply and measure currents and voltages through the sample. Electrical contact from the Hall/Van der Pauw controller to the specimen was accomplished by cold-soldering indium-coated platinum wires to the electron beam evaporated contacts. Data was acquired with a custom LabView program.

8.3.2 3. Results and Discussion

8.3.3 3.1 Ultraviolet photoelectron spectroscopy (UPS) and X-ray photoelectron spectroscopy

(XPS)

Ultraviolet photoelectron spectroscopy was used to measure the work functions of the surfaces of the CGA sample and the Au contact and these results are plotted in Figure 2. An accelerating potential was used to separate the secondary edges of the sample and analyzer, which allows for an accurate determination of the position of the secondary edge of the sample, and hence the work function. In the UPS spectrum, the work function is the difference between the energy of the ultraviolet photons (hν = 21.2 eV for He I radiation) and Eb, the binding energy of the secondary edge. The secondary edge is the maximum binding energy found for photons in the spectrum and represents the minimum kinetic energy for photoelectron escape from the surface. The location of the secondary edge is fit experimentally in various ways in practice;13,14 we chose to fit the edge with a Gaussian-Lorentzian peak and record the location of the secondary edge to be one half-width at half maximum toward the higher binding energy side of

8.30

the edge. This gave values of φAu = 5.15 eV and φCGA = 4.37 eV, a difference of 780 meV. The measured work function value we obtained for Au is in excellent agreement with a value obtained by KPFM on a similar Au sample of φAu = 5.2 eV measured relative to SrTiO3 and

15 Cu2O.

With XPS, we determined that the valence band edge in the CGA was offset by +0.53 eV with respect to the Fermi level of the Au. Based on the measured work functions from UPS, we can calculate that the Fermi level of the CGA sits 0.25 eV above its valence band edge. This indicated that the CGA sample was p-type, in agreement with hot point probe measurements.

The mobility gap in CGA was characterized previously by optical absorption spectroscopy and ellipsometry.12 The onset of the mobility edge above the background subgap absorption (for samples with the same composition used in this experiment) began at 0.65 eV and appeared to converge to a band-like behavior at approximately 1.1 eV.12 With information from UPS, XPS and the mobility gap, the electron affinity of the CGA was calculated to be 3.52 eV.

The Mg and Al contacts were not tested in the UPS or XPS because their work functions vary with surface conditions much less than for Au, and so the literature values for work function, 3.66 eV and 4.06-4.41 eV, respectively, were used. Based on the electron affinity of

CGA, calculated above, we conclude that we should have obtained an ohmic contact between Au and CGA and a Schottky contact with a contact barrier height of ~700 meV between the CGA and the Mg. A summary of measured and calculated values is shown in Figure 3.

8.31

8.3.4 3.2 Kelvin probe force microscopy

The sample and associated contact metals were measured by KPFM. Many factors can influence the KPFM-measured contact potential difference (CPD) and lead to a change in work function at a surface, including adsorption of gas atoms, relaxation or reconstruction of surfaces, differences in crystallographic faces, built up charge on the surface, surface photovoltage, induced surface-dipoles, and tip conditions.16 As a result, comparisons within individual KPFM measurements are more reliable than comparisons between two different sample measurements.

Thus the current study examined metal layers directly on a CGA sample, with measurements across the metal-CGA junction.

Areas that included relatively abrupt edges of both contact materials were analyzed with

AFM and KPFM. Figures 4(a) and (b) present surface topography (greyscale) and contact potential difference (CPD) (color overlay) for the two contacts studied. Au/CGA edges were typically much sharper than Al/Mg/CGA edges in both topography and contact potential difference. Away from the contact edge, the roughness of the bare CGA and metal contact are significant with respect to the contact edge thickness itself; however, the CPD signal did not vary significantly with topography away from the contact edge. The minimum CPD on these 3D images is set to zero by the MFP3D software, so the CPD is a relative value for each scan.

Figures 4(c) and (d) give corresponding linescans across metal/CGA edges based on the data in

Figures 4(a) and (b). The linescans are averages of approximately 30 neighboring traces spaced

0.1 µm apart.

Table I gives the relevant CPD values and contact edge height differences, Δz, along with measured values from UPS and XPS. Each CPD is an areal root-mean squared average over a 3-

6 µm2 portion of a 15 × 15 µm KPFM scan. We found that the CPD of the CGA substrate varied

8.32

with location. It depended on proximity to a metal contact and on the type of metal contact that was neighboring the CGA during the CPD measurement. Specifically, the CPD for CGA varied from as low as 404 ± 17 mV near Au contacts to as high as 818 ± 11 mV near Al/Mg contacts.

CPD values for bare CGA, not near any contact metal (at least 1000 µm from any contact) were in the range of 571 ± 12 mV. Therefore, two ΔCPD values are reported, one for the change in

CPD across the edge in the immediate vicinty of the contact (ΔCPDedge) and one for the difference between the metal CPD and the bare, isolated CGA CPD (ΔCPDlong range). We suggest that ΔCPDlong range represents the real work function differences between the metal contacts and the CGA, as opposed to ΔCPDedge which includes the local influence of the metal.

For the Au contact, the CPD decreased by an average of ΔCPDlong range = 500 mV as the tip went from bare CGA to Au. This measurement is considerably lower than the work function difference of 780 meV measured by UPS but is within experimental error, due in part to air exposure of the surfaces studied by KPFM and due to sputter cleaning of the surfaces studied by

UPS/XPS. Despite these uncertainties, we can conclude that work function of the CGA is in the range of ~450-800 meV less than that of Au, and given this we should expect an ohmic contact.

The corresponding contact band edge diagram is shown schematically in Figure 3.

For the Al/Mg contact, the CPD increased by an average of ΔCPDlong range = 410 mV as the tip went from bare CGA to Al. We have two ways to calculate φAl. Using the UPS- measured φCGA = 4.37 eV, we calculate that φAl = 4.37-0.41 = 3.96 eV. The KPFM measurements for Δφ between Au and Al was 910 meV, yielding φAl = 4.24 eV. The average of the two values for φAl from the KPFM here are in good agreement with the literature range of

4.06-4.41 eV for Al. The CGA is in contact with Mg (φ = 3.66 eV) and not the exposed Al capping layer, measured in KPFM. Therefore we expect that the Mg/CGA would have a ΔCPD

8.33

0.3 to 0.6 eV greater than does the Al/CGA interface (see schematic band edge diagrams in

Figure 3). Therefore we expect φMg to be ~0.7 eV below φCGA. In any case, we expect both Al and Mg to produce Schottky barrier contacts, although the Mg/CGA barrier height should have been much higher.

The morphological widths of the interfaces were 0.32 ± 0.15 µm for the Au contact and

2.60 ± 0.15 µm for the Al/Mg contact as judged from the point where the metal/CGA edge began to decrease down to the point where the topography flattened out on the CGA side. This longer interfacial width for the Al/Mg contact was most likely due to lateral surface self-diffusion of the metal layers under the TEM grid mask during electron beam evaporation. Atoms arriving from the vapor phase provide thermal energy to the metal surface (50-60°C) during the deposition, allowing for modest surface self-diffusion. In general, the surface diffusion coefficient of metals increases with increasing homologous temperature (TH), the ratio of the temperature of a metal surface to its melting point.17 Therefore, surface self-diffusion is expected to be lower for Au

(TH = 0.24) than for Mg (TH = 0.36) or Al (TH = 0.35).

The width of the interface in terms of the potential was found in a similar way by plotting the derivative of the potential with respect to distance (not shown). The CPD interface width was 1.96 ± 0.2 µm for the Au contact and 5.4 ± 1.0 µm for the Al/Mg contact. The inflection points for CPD and topography in the derivative plots occur at the same position for both contact types, indicating that the CPD change is directly related to the contact edge location.

It is possible that the width, w, of carrier depletion around a metal/CGA interface extended laterally into the volume of CGA material and was responsible for the more gradual change in CPD as compared to the change in topography across the contact edge. The expected depletion width in a metal/semiconductor junction can be estimated by using a p+-n

8.34

approximation; that is, by assuming the charge in the metal is sufficient to produce essentially no depletion in the metal and that the carrier concentration in the metal is much greater than that in the CGA, which we have measured to be very small as discussed above.18 In this case the depletion width extends only into the semiconductor side of the junction and is described by

2εV w = 0 , qN a where ε is the permittivity of the CGA, V0 is the contact potential, and Na is the acceptor concentration in the p-type CGA. V0 is directly measured in the KPFM; it is the difference in

13 -3 work functions between the metal and CGA. Na is estimated to be 6 × 10 cm from room

5 temperature resistivity data of the CGA (~10 Ω-cm) and an estimate of its hole mobility (µp~1

2 19 cm /Vs) taking that of a-Si as roughly comparable. The permitivity, ε, is projected to be 15ε0.

This estimation led to values of w = 3.6 µm for Au contacts and w = 4.9 µm for Al/Mg contacts.

This calculation predicts that, regardless of the particular values of semiconductor parameters chosen, the Al/Mg contact produces a wider depletion width than the Au contact by the factor

w V Mg / Al ∝ 0,Mg / Al . wAu V0,Au

The modeled ratio is 1.4 using V0 as measured by KPFM. The measured ratio from analysis of linescans was 5.4/1.96 = 2.76 (but within error can range as low as 2.04). Although rudimentary, this approximation shows that the measured CPD interface width in KPFM agrees at least qualitatively with what would be expected for a metal/CGA depletion width extending laterally into the CGA material. Note that if the shape of the CPD change across the metal/CGA edge were due entirely to a p+-n depletion width as described above, we would expect to observe completely asymmetric potential spreading, only on the CGA side. However, the KPFM data show some degree of symmetry, although the CPD change is less sharp on the CGA side for both

8.35

contact types. This is likely because the CPD is a superposition of the effect of topography

(giving rise to symmetry) and the depletion into the CGA side (giving rise to asymmetry).

3.3 Electrical Measurements

Electrical measurements were performed on a number of metal/CGA combinations including CGA samples with a wide range of Ge contents (CuGexAs2 with x=0.45, 0.65, 0.85, and 1.00). Here, we describe in detail those measurements for the KPFM-analyzed specimen.

The other metal/CGA combinations behaved similarly.

Four-point probe measurements on the CGA samples were used to characterize the conductivity as a function of temperature, as shown in Figure 5. The data fit well with a

Boltzmann distribution resulting in an activation energy for conduction in the sample of 0.52 eV, consistent with activation of carriers from midgap states into the corresponding bands.

Temperature dependent current-voltage (I-V) curves were measured for three specific contact configurations: between two Au contacts, between two Al/Mg contacts, and from one type of contact to the other. Representative log-I-V curves at various temperatures are shown in

Figure 6 for the Al/Mg-CGA-Au contact configuration. As the sample temperature dropped below 220 K, resistivities became so large that our system was not able to drive measurable current through the samples. For the three contact configurations, I-V curves were linear in both directions of current flow, showing no rectifying behavior throughout the temperature range measured. Arrhenius plots were made of conductivity vs. inverse temperature for each contact configuration, and were fitted with an exponential function to calculate the activation energy for electrical conduction (Table II). These values were in reasonable agreement with the activation energy determined from the four-point probe measurement. This indicates that the resistance of

8.36

the conduction path for all contact combinations was mostly due to the CGA itself and not due to either of the contact types.

Based on the sample geometry, the expected series resistance of a given current path can be calculated. Furthermore the contact resistances based on the standard Schottky diode thermionic emission model can be estimated to determine if the resistance should be dominated by the expected Schottky barriers.20 For a Schottky diode the current I at bias voltage V is given by

qV /nkT I = I0 (e !1) where

I A* AT 2 exp( q / k T) 0 = − Φ B ,

* and A is the effective Richardson constant, A is the diode area, kBT/q is the thermal voltage, n is the diode ideality factor and Φ is the Schottky barrier height. Ay et al.21 report experimental

Richardson constants in a wide range of 0.341-438 A/m2K2 for a-Si:H/Au Schottky barriers.

Hull et al.22 report A* = 782 A/m2K2 for c-Ge and a-Ge Schottky barriers. Here we chose values of A*=100 and 500 A/m2K2 as representative values. The resulting estimates of contact properties for the KPFM-measured Schottky barrier heights are given in Table III, for given T, Φ and A*. Table 3 gives the reverse voltage (Vt) at which the series resistance of the CGA material matches that of an ideal Schottky barrier in reverse bias. Since every value of Vt is less than 1 V, we conclude that the voltage range used in this experiment (± 2 V) is adequate to measure the possible effect of a Schottky barrier between CGA and Mg (or Au). Variation of A* within the range reported for other amorphous semiconductor-metal pairs does not change this conclusion.

A non-ideal diode (n=2) would require more voltage to drive a given current through the

Schottky diode so non-ideal behavior does not account for the difference. Although a Schottky contact would be expected for the Mg contacts on p-type CGA, the contacts made in this

8.37

experiment were not Schottky-like or were highly imperfect and did not contribute significantly to impeding the current when compared with the resistance of the CGA material itself. Other contact metals were also tested, including Ni, Al, and Pt, for each of the CuGexAs2 alloy compositions x=0.45, 0.65, 0.85 and x=1, yet nothing was found that yielded a Schottky contact on amorphous CGA at temperatures from 300 K to as low as 220 K. The exception was one set of contacts measured between two sets of Mg/Al contacts, which should have appeared resistive.

Therefore we conclude that measurement was affected by local non-uniformities in the CGA in that sample that were undetectable by the methods used here.

4. Conclusions

Kelvin probe force microscopy measurements of Au and Al/Mg contacts to a bulk amorphous CGA sample show contact potential differences consistent with the work function difference of the metals. UPS was used to measure the work functions of Au φAu = 5.15 eV and of the CGA φCGA = 4.37 eV. XPS was used to measure the valence band offset of the CGA with respect to the Fermi level of Au. These measurements led to a schematic band diagram of the

CGA in contact with the metals chosen. We conclude that work function of the CGA is in the range of ~450-800 meV less than that of Au, and given this we should expect an ohmic contact.

The electron affinity of the CGA was calculated to be 3.52 eV. A work function difference of 0.70 eV for CGA/Mg contacts was measured, which we expected to result in a

Schottky contact. The behavior of the CPD around the edges of the contacts was interpreted in terms of the formation of a depletion region around the contact and diffusion of Al and Mg atoms during deposition, but the region affected by the metal appears to extend far beyond the range of normal depletion. The CPD from the CGA to the metals puts the work function of this

CGA sample rougly midway between the two metals. Given the work function difference and

8.38

the mobility gap in the CGA, the Mg/Al contacts should have yielded a relatively high Schottky barrier and a highly resistive contact for one bias direction. The Au contact should have yielded an ohmic contact. However, electrical measurements show no diode-like behavior for any combination of contacts measured. The resistance of the circuit from any contact to any other was consistent with the resistance of the CGA material itself at the measurement temperature.

The temperature dependencies of CGA with different electrode contact metals also followed the same behavior as CGA without an electrode. We conclude that these materials did not produce a

Schottky contact on the CGA sample studied despite contact potential differences that should have yielded a good Schottky contact. This result is not specific to the sample or metals measured and we have not obtained a reproducible Schottky contact on any CGA sample using any metal.

Although we could not measure any local variations in conduction in the CGA as a function of position we propose that local inhomogeneities on a scale less than what could be measured with the KPFM or XPS/UPS led to local ohmic contacts to the metals in all cases.

This could be the case based on local short-range order suggested by the XRD. Any larger patch of material with high conductivity could have yielded an ohmic contact behavior if the area of those patches were sufficient to carry the currents used here. However, we note that if these areas were common enough to cause all contacts to appear ohmic we would have expected to observe them with the KPFM. If they are common but too small to be distinguished with the

KPFM we should have observed variations in potential across the surface or very little depletion region. If there were an interfacial reaction under the contact that was unobservable with the techniques used here we would have expected one of the various metals on one of the various

8.39

compound compositions to yield a diode-like behavior. Therefore we are unable to explain the lack of diode-like current voltage curves in our measurements.

8.4 Acknowledgements

This work was supported by the U.S. Department of Energy, Office of Nonproliferation

Research and Development (NA-22). Pacific Northwest National Laboratory is operated for the

U.S. Department of Energy by Battelle Memorial Institute under Contract No. DE-AC05-

76RLO1830. Materials analysis was carried out in the Frederick Seitz Materials Research

Laboratory Central Facilities at the University of Illinois which are partially supported by the

College of Engineering. The authors thank Scott MacLaren for his work in scanning probe microscopy, the group of Dr. Jim Eckstein for lending cryogenic wiring and varnish, Dr. Julio

Soares for helping with the setup of the I-V curve station and LabView programming, and Laura

Burka and Clyde Chamberlin for their help in preparing the samples.

8.40

Table I. KPFM and UPS/XPS data summary

KPFM UPS XPS Δφ barrier Contac CPD ΔCPD ΔCPD ΔVBE Δz (nm) area edge long height t edge (meV) (meV) (meV) (meV) range (meV) 72.0±1. CGA 404±17 Au -333±24 -500 780 530 1 Au 71±24 CGA 818±11 Al/Mg 75±12 166±17 413 700* -- Al 984±17 bare CGA 571±12

Table II. Activation energies for three conduction paths (contact configurations)

Contacts Ea (eV) Au – Au 0.517 Al/Mg – Al/Mg 0.461 Au – Al/Mg 0.510

Table III. Estimates of contact properties for KPFM-measured Schottky barrier heights

A* = 100 A* = 500 2 2 2 2 A/m K A/m K

T (K) Rs (Ω) Φ (eV) Vt (V) Vt (V) 0.1 -0.650 -0.688 300 1.E+06 0.6 -0.150 -0.188 0.1 -0.620 -0.652 250 1.E+08 0.6 -0.120 -0.152 0.1 -0.588 -0.616 220 2.E+09 0.6 -0.088 -0.116

8.41

Figure/Table Captions

Figure 1. Squares of Al/Mg contact metal (58 µm x 58 µm, bar width 25 µm, thickness ~70 nm) electron- beam deposited through a TEM grid on a bulk amorphous CGA substrate.

Figure 2. UPS data for Au and CGA material. The secondary edge fit is illustrated.

Figure 3. Band diagrams of metal/CGA (a) before and (b) after contact, based on UPS, XPS, KPFM, optical absorption spectroscopy data. Values are in eV.

Figure 4. (color online) Kelvin Probe Force Microscopy maps of contact potential difference (CPD) on the surface of (a) Au contact/CGA edge and (b) Al/Mg contact/CGA edge. The height data represent the topography of the surface while the color represents the CPD value. For the Au contact, the right-hand side has large CPD; for the Al/Mg contact, the left-hand side has large CPD. The white lines on the KPFM maps indicate the regions used for corresponding linescans across metal/CGA edge, shown in (c) and (d). Profiles shown are averages of approximately 30 neighboring linescans spaced at 0.1 µm. Linescans were measured perpendicular to the contact/CGA edge.

Figure 5. Conductivity vs. temperature plot for CGA sample in four point probe configuration.

Activation energy Ea is calculated by fitting the curve to an exponential function.

Figure 6. Temperature dependent I-V curves for contact configuration Al/Mg-CGA-Au. I-V curves are linear and symmetrical. Curves were taken every 2 K; curves shown here are for every 10 K.

Table I. *Indirectly calculated value.

Table II.

Table III. Comparison of the reverse voltage (Vt) at which a Schottky barrier with the given parameters

(A*, barrier height Φ, T) provides the same resistance as the series resistance Rs for the CGA material itself.

8.42

References

1 L. Bai, N. Y. Garces, C. Xu, L. E. Halliburton, N. C. Giles, P. G. Schunemann, K. Nagashio, C. Yang, and R. S. Feigelson, in Effect of donors and acceptors on the optical properties of

CdGeAs2, San Jose, CA, USA, 2004 (SPIE), p. 22. 2 V. Rud, Y. Rud, I. Polushina, T. Ushakova, and S. Iida, "Observation of Record Hall Mobility in

CdGeAs2 Single Crystals," Japanese Journal of Applied Physics 39, 266 (2000). 3 S. H. Risbud, "Processing and properties of some II-IV-V-2 amorphous and crystalline semiconductors," Applied Physics A-Materials Science & Processing 62, 519 (1996). 4 U. Rau and H. W. Schock, "Electronic properties of Cu(In,Ga)Se2 heterojunction solar cells - recent achievements, current understanding, and future challenges," Applied Physics A-Materials Science & Processing 69, 131 (1999). 5 N. Dietz and F. L. Madarasz, "Chemical and biological sensors based on optically confined birefringent chalcopyrite heterostructures," Materials Science and Engineering B-Solid State Materials for Advanced Technology 97, 182 (2003). 6 M. D. Pelusi, V. G. Ta'eed, L. B. Fu, E. Magi, M. R. E. Lamont, S. Madden, D. Y. Choi, D. A. P. Bulla, B. Luther-Davies, and B. J. Eggleton, "Applications of highly-nonlinear chalcogenide glass devices tailored for high-speed all-optical signal processing," IEEE Journal of Selected Topics in Quantum Electronics 14, 529 (2008). 7 S. Sadewasser, "Surface potential of chalcopyrite films measured by KPFM," Physica Status Solidi A-Applications and Materials Science 203, 2571 (2006). 8 C. Lian and H. L. Xing, "Surface potential measurements on Ni-(Al)GaN lateral Schottky junction using scanning Kelvin probe microscopy," Applied Physics Letters 88 (2006). 9 S. Sabuktagin, Y. T. Moon, S. Dogan, A. A. Baski, and H. Morkoc, "Observation of surface charging at the edge of a Schottky contact," IEEE Electron Device Letters 27, 211 (2006). 10 E. Schlenker, V. Mertens, J. Parisi, R. Reineke-Koch, and M. Kontges, "Schottky contact analysis of photovoltaic chalcopyrite thin film absorbers," Physics Letters A 362, 229 (2007). 11 H. B. Michaelson, "Work function of elements and its periodicity," Journal of Applied Physics 48, 4729 (1977). 12 B. R. Johnson, B. J. Riley, S. K. Sundaram, J. V. Crum, C. H. Henager Jr., Y. Zhang, V. Shutthanandan, C. E. Seifert, R. M. Van Ginhoven, C. E. Chamberlin, A. A. Rockett, D. N. Hebert, and A. R. Aquino, "Synthesis and Characterization of Bulk, Vitreous Cadmium Germanium Arsenide," Journal of the American Ceramic Society 92, 1236 (2009). 13 D. Briggs and J. C. Riviere, Practical Surface Analysis, Vol. 1, 2nd ed. (John Wiley and Sons, Chichester, England, 1983). 14 S. Hufner, Photoelectron Spectroscopy Principles and Applications (Springer, Heidelberg, 1995). 15 S. Lee, C. W. Liang, and L. W. Martin, "Synthesis, Control, and Characterization of Surface

Properties of Cu2O Nanostructures," ACS Nano 5, 3736 (2011). 16 E. H. Rhoderick and R. H. Williams, Metal-semiconductor contacts, 2nd ed. (Claredon Press; Oxford University Press, New York, 1988). 17 G. Neumann and G. M. Neumann, Surface self-diffusion of metals (Diffusion Information Center, Bay Village, Ohio, 1972). 18 B. G. Streetman, Solid state electronic devices, 4th ed. (Prentice Hall, Englewood Cliffs, N.J., 1995). 19 C. Xu, Thesis, West Virginia University, 2007.

8.43

20 L. Kronik and Y. Shapira, "Surface photovoltage phenomena: theory, experiment, and applications," Surface Science Reports 37, 1 (1999). 21 I. Ay and H. Tolunay, "The influence of ohmic back contacts on the properties of a-Si:H Schottky diodes," Solid-State Electronics 51, 381 (2007). 22 E. L. Hull and R. H. Pehl, "Amorphous germanium contacts on germanium detectors," Nuclear Instruments & Methods in Physics Research Section A - Accelerators Spectrometers Detectors and Associated Equipment 538, 651 (2005).

8.44